Merge pull request 'next into main' (#110) from next into main

Reviewed-on: #110

README.md

@@ -3,22 +3,22 @@
 ## Description
 
 `sfdupes` is an MIT-licensed Go CLI tool by [@sneak](https://sneak.berlin) that
-quickly finds _candidate_ duplicate files, and ultimately whole duplicate
+quickly identifies _candidate_ duplicate files — and, ultimately, entire
-directory trees, on very large filesystems. It does this without reading every
+duplicate directory trees — across very large filesystems without reading every
-byte of every file.
+byte of every file. Files are considered duplicates when their sizes are equal
-
+and they agree on a short ladder of hashes. A file under 10 MiB is hashed in
-Two files are duplicates when they have the same size and the same three SHA-256
+full and compared directly. A larger file is gated first on the SHA-256 of its
-hashes. Two of the hashes cover the first and the last 64 KiB of the file. The
+first 64 KiB and of its last 64 KiB, and only when its size and both of those
-third, the content hash, covers the whole file when it is under 100 MiB, so a
+match another file's is it read for a content hash to compare — the SHA-256 of
-match proves identical content. At 100 MiB and above it covers only 1 MiB at
+the whole file when it is under 50 MiB, or of gigabyte-spaced 1 MiB samples when
-each gigabyte, so such a file is never read in full. A match there is a strong
+it is 50 MiB or larger. Below 50 MiB the content hash is proof of identical
-candidate rather than proof.
+content; at or above 50 MiB it is a strong candidate signal rather than proof,
-
+because the gaps between samples are never read. The intended use is finding
-The intended use is finding duplicate downloads and duplicated directory trees
+duplicate downloads and duplicated directory trees on multi-terabyte ZFS servers
-on multi-terabyte ZFS servers, where reading every byte of every file is
+where reading every byte of every file is prohibitively expensive. `scan`
-prohibitively expensive. `scan` keeps a persistent SQLite database of file
+maintains a persistent SQLite database of file signatures that survives between
-signatures between runs, so it can run from cron. The reports can be generated
+runs, so it can be run from cron and the reports can be generated at any time
-at any time from the most recent scan.
+from the most recent scan.
 
 This README is the complete and authoritative specification.
 
@@ -34,13 +34,13 @@ export SFDUPES_DATABASE="$HOME/.local/share/sfdupes/db.sqlite"
 
 `scan` walks one or more filesystem trees and maintains one database record per
 regular file (path, size, mtime, head hash, tail hash, content hash). The
-database persists between runs. A rescan hashes only the files that are new or
+database persists between runs; a rescan only hashes files that are new or
-changed, and removes the records of files that no longer exist. `report` reads
+changed, or that may have gained a duplicate since the last scan, and removes
-the database and prints the file-level duplicates report. `trees` reads the same
+records for files that no longer exist. `report` reads the database and prints
-database and prints the duplicate-tree report. `info` prints the database's
+the file-level duplicates report. `trees` reads the same database and prints the
-details: its path, the sizes of its files and the number of its records. A
+duplicate-tree report. `info` prints the database's details: its path, the sizes
-missing/invalid subcommand — or a `scan` invocation with no `PATH` operand —
+of its files and the number of its records. A missing/invalid subcommand — or a
-prints a usage message and exits 2.
+`scan` invocation with no `PATH` operand — prints a usage message and exits 2.
 
 The database defaults to `/var/lib/sfdupes/db.sqlite` and can be placed anywhere
 by setting `SFDUPES_DATABASE`. The intended deployment is a daily `sfdupes scan`
@@ -122,19 +122,14 @@ As a file `/etc/cron.d/sfdupes`:
   the first scan. To keep it elsewhere, set
   `SFDUPES_DATABASE=/path/to/db.sqlite` before the command on the same line.
 - `scan` writes nothing to stdout. Its stderr, appended here to
-  `/var/log/sfdupes.log`, holds these lines:
+  `/var/log/sfdupes.log`, holds a first line naming the database (see
-    - a first line naming the database (see "Database");
+  "Database"), plain progress lines naming each phase and its place among the
-    - plain progress lines naming each phase and its place among the four, as
+  five, as `[2/5] walk` (one as the phase starts, one every 5 seconds while it
-      `[2/4] walk`: one as the phase starts, one every 5 seconds while it runs,
+  runs, and one as it ends), a warning for each path it skips, a line for each
-      and one as it ends;
+  default exclude its walk does not enter and for each path it leaves out
-    - a warning for each path it skips;
+  because it is not on a persistent filesystem (see "Rules for the walk"), and
-    - a line for each default exclude its walk does not enter, and for each path
+  the summary line (see "Progress" and "`scan` mode"). The log grows with every
-      it leaves out because it is not on a persistent filesystem (see "Rules for
+  scan; rotate it like any other.
-      the walk");
-    - the summary line (see "Progress" and "`scan` mode").
-
-    The log grows with every scan; rotate it like any other.
-
 - Skipped paths do not fail a scan: it still exits 0, and cron sends nothing. A
   scan that fails, or is stopped by `SIGINT` or `SIGTERM`, exits 1 with the
   reason among the last lines of the log; `tail` prints them, and cron mails
@@ -158,7 +153,7 @@ A row is a candidate, not proof:
 
 - The reports read only the database, so they show the files as of the last
   scan; a file may have changed or gone since.
-- A file of 100 MiB or more is compared only on samples of its content (see
+- A file of 50 MiB or more is compared only on samples of its content (see
   "Duplicate detection").
 - Paths that are hard links to one file are listed as duplicates, but they share
   their data, so removing one frees nothing.
@@ -193,23 +188,22 @@ or a symlink, which `trees` does not see.
 
 ## Rationale
 
-Duplicate finders that hash every file in full do not scale to the target
+Duplicate finders that hash entire files do not scale to the target environment:
-environment. It holds about 10 million files and 150 TB, on disks that may be
+~10 million files and ~150 TB on possibly slow or busy disks (a ZFS pool under
-slow or busy, such as a ZFS pool under resilver.
+resilver). sfdupes spends disk I/O only on files whose size at least one other
-
+file shares, since a size-unique file cannot be a duplicate. Of those, a file
-sfdupes reads a file under 100 MiB in full, and a larger one only at its two
+under 10 MiB is read in full; a larger one has its cheap end windows read first,
-ends and at 1 MiB per gigabyte. A full read of a small file is cheap, and it
+and is read for a content hash only when its size and both end windows match
-makes the comparison exact. A full read of every large file would take
+another file's — the whole file below 50 MiB, but only gigabyte-spaced samples
-unreasonably long. Sampling cuts a 150 GiB file to about 150 MiB of reads, at
+at or above 50 MiB, so the largest files are never read in full. This keeps a
-the price of a comparison that is not exact.
+full-filesystem sweep tractable, and the signatures are kept in a persistent
-
+database, so the expensive filesystem pass is incremental: a rescan re-hashes
-The hashes are kept in a persistent database, so the expensive filesystem pass
+only files whose recorded mtime or size changed, plus — for its content hash — a
-is incremental. A rescan reads only the files that are new or whose size or
+file of 10 MiB or more whose size and end windows have come to match another
-mtime changed. All analysis then works from the database alone.
+file's. All analysis happens offline from the database alone. The end goal is
-
+not individual files but whole duplicated trees — duplicate extractions,
-The end goal is not individual files but whole duplicated trees, such as a
+duplicate downloads, copied project trees — which an operator can consider
-download extracted twice. An operator can consider removing such a tree as a
+removing as a unit.
-unit.
 
 ## Design
 
@@ -221,10 +215,14 @@ Goals, in order:
     trees), so the operator can consider removing an entire subtree at once.
     File-level duplicate detection is the foundation; tree-level detection is
     built on top of it.
-2.  **Bound the I/O per file.** A file is read only when it is new or has
+2.  **Spend I/O in proportion to duplicate likelihood.** Only files whose size
-    changed. A file under 100 MiB is then read in full, and a larger one only at
+    at least one other file shares are read at all — a size-unique file cannot
-    its ends and at 1 MiB per gigabyte (see "Duplicate detection" below).
+    be a duplicate. Those are compared by the ladder in "Duplicate detection"
-    Therefore no file of 100 MiB or more is read in full. Scale target: tens of
+    below: a file under 10 MiB is hashed in full, while a larger file is gated
+    on cheap 64 KiB end windows first, and gets a content hash only when its
+    size and both end windows match another file's. That hash reads the whole
+    file below 50 MiB but only gigabyte-spaced 1 MiB samples at or above it, so
+    the very largest files are still never read in full. Scale target: tens of
     millions of files, ~150 TB filesystem, possibly slow or busy disks (ZFS pool
     under resilver). Holding one small record (path, size, mtime) per file in
     memory during a scan is acceptable; holding every file's hashes is not (they
@@ -237,10 +235,12 @@ Goals, in order:
     being read, so its memory grows with the number of directories and with the
     size of the largest directory.
 3.  **Scan incrementally, analyze offline.** The expensive filesystem scan
-    maintains a persistent database, and an unchanged file is never read again
+    maintains a persistent database; an unchanged file is never read again on a
-    on a rescan. All analysis (`report`, `trees`) works from the database alone
+    rescan, except to compute its content hash once a file of 10 MiB or more
-    and must never touch the scanned filesystem again. `scan` is designed to be
+    comes to match another on size and both end windows. All analysis (`report`,
-    cronned; the reports run at any time against the last completed scan.
+    `trees`) works from the database alone and must never touch the scanned
+    filesystem again. `scan` is designed to be cronned; the reports run at any
+    time against the last completed scan.
 4.  **Clean stream separation.** Everything on stdout is machine-readable data.
     All progress, warnings, summaries, and help and usage text go to stderr.
     Never mix them.
@@ -272,7 +272,7 @@ Four subcommands, all implemented:
     from the database records, compute a Merkle-style digest per directory, and
     report maximal groups of identical trees.
 4.  `info` — the database's details: its path, the sizes of its files, its
-    schema version, the number of its records and their total size.
+    schema version and the counts of its records.
 
 ```
 sfdupes scan [--workers N] [-x] [--no-default-excludes] PATH...
@@ -368,7 +368,9 @@ All four subcommands operate on a single SQLite database file:
     - While it loads the existing records (`load`) and walks: no records. A
       first scan's database holds none, however long its walk has run.
     - During the hash phase: the first records, committed 10,000 at a time.
-    - In the update phase: the rest of the hashed records, and the deletions.
+    - In the update phase: the rest of the hashed records, the records for
+      size-unique files, and the deletions.
+    - In the content phase: content hashes, again committed 10,000 at a time.
     - At the end, as it closes the database: the `-wal` and `-shm` files go
       away, unless a report still has the database open (see above).
 - `report`, `trees` and `info` open the database read-only and need only read
@@ -388,9 +390,9 @@ All four subcommands operate on a single SQLite database file:
         size       INTEGER NOT NULL,  -- bytes, from lstat
         mtime      INTEGER NOT NULL,  -- whole Unix seconds of the mtime, from lstat
         mtime_nsec INTEGER NOT NULL,  -- nanoseconds within that second, 0 to 999999999
-        head       BLOB NOT NULL,     -- raw 32-byte SHA-256 of the first 64 KiB
+        head       TEXT NOT NULL,     -- lowercase-hex SHA-256; first 64 KiB, or whole file under 10 MiB
-        tail       BLOB NOT NULL,     -- raw 32-byte SHA-256 of the last 64 KiB
+        tail       TEXT NOT NULL,     -- lowercase-hex SHA-256; last 64 KiB, or whole file under 10 MiB
-        content    BLOB NOT NULL      -- raw 32-byte SHA-256; whole file under 100 MiB, else samples
+        content    TEXT NOT NULL      -- lowercase-hex SHA-256, whole file or samples
     ) WITHOUT ROWID;
     CREATE INDEX files_signature ON files (size, head, tail, content);
     ```
@@ -401,46 +403,57 @@ All four subcommands operate on a single SQLite database file:
     nanoseconds past that second. Split this way they hold any mtime a
     filesystem can record, one before 1678 or after 2262 included, which a
     single 64-bit count of nanoseconds cannot. They are used only for change
-    detection and are not part of the duplicate key. Every record holds the
+    detection and are not part of the duplicate key. For a file under 10 MiB
-    three hashes that "Duplicate detection" below describes. The
+    `head`, `tail`, and `content` all hold the whole-file hash (that range is
-    `files_signature` index lets `report` read the records in signature order
+    hashed in full, with no end windows); for a larger file `head` and `tail`
-    without sorting the table.
+    hold the first- and last-64 KiB hashes and `content` the whole-file or
+    sampled hash. All three are empty strings when the file has never been
+    hashed because its size was unique as of the last scan that covered it. For
+    a file of 10 MiB or more, `content` stays empty until the content phase of a
+    scan (see "`scan` mode" below) has read the file. A record with an empty
+    `content` is never part of a duplicate group, though it still defines the
+    file for tree reconstruction. The `files_signature` index lets `report` read
+    the records in signature order without sorting the table.
 
 ### Duplicate detection
 
-Two files are duplicates when they have the same size and the same three hashes.
+Two files are duplicates only when they agree on every rung of this ladder; a
-`scan` computes the hashes and stores them in the database. `report` and `trees`
+mismatch at any rung means they are not duplicates. `scan` stores each file's
-group the files whose size and hashes all match, whichever scan recorded them.
+hashes, and `report` and `trees` group files by the whole signature — size,
+`head`, `tail`, and `content` — so the grouping is exactly this ladder applied
+across everything scanned into the database, even across separate scans.
 
-`scan` gives every file it records three SHA-256 hashes:
+1.  **Size.** Files of different sizes are never compared. Only files whose size
+    at least one other file shares are hashed at all.
+2.  **Under 10 MiB: whole file.** A file smaller than 10 MiB is hashed in full
+    and compared directly, with no separate end-window step — small files are
+    cheap to read to the last byte, and doing so makes the comparison exact.
+    `head`, `tail`, and `content` all hold this whole-file SHA-256, so such a
+    file's signature is decided entirely by its size and its content.
+3.  **10 MiB and above: head and tail.** For a larger file, the SHA-256 of the
+    first 64 KiB (`head`) and of the last 64 KiB (`tail`) are a cheap gate that
+    eliminates most same-size pairs before any bulk reading: the content hash of
+    the next two rungs is computed only for a file whose size, `head`, and
+    `tail` match another file's, whether that file is scanned in the same run or
+    stored by an earlier scan. A stored file that first gains such a match in a
+    later scan gets its content hash then; until it has one, its `content` is
+    empty and it is not a duplicate. At 10 MiB and above the two windows never
+    overlap.
+4.  **10 MiB and above, content below 50 MiB.** The SHA-256 of the entire file.
+    Agreement here is proof of identical content (barring a SHA-256 collision).
+5.  **10 MiB and above, content 50 MiB and above.** A sampled SHA-256: the 1 MiB
+    window at each gigabyte-aligned offset (0, 1 GiB, 2 GiB, … while inside the
+    file, the final window truncated at end of file) is fed, in order, into one
+    hash. This is **deliberately probabilistic** — the gaps between samples are
+    never read, so two large files that agree on every sample are reported as
+    duplicates without being read in full. It is the price of never reading a
+    150 GB file end to end. Because size is already part of the signature, only
+    equal-size files reach this rung, so their sample boundaries always align.
 
-- `head`, the hash of the file's first 64 KiB.
+`head`, `tail`, and `content` are one column each. A file below 10 MiB and one
-- `tail`, the hash of the file's last 64 KiB.
+at or above it never share a size, and neither do a file below 50 MiB and one at
-- `content`, the content hash described below.
+or above it, so a stored value is never ambiguous between the whole-file,
-
+end-window, and sampled forms.
-In a file of 64 KiB or less, the first and the last 64 KiB are both the whole
-file, so all three hashes are the hash of the whole file.
-
-The content hash depends on the file's size:
-
-- Under 100 MiB, it is the SHA-256 of the whole file. Agreement on it is proof
-  of identical content, barring a SHA-256 collision.
-- At 100 MiB and above, it is the SHA-256 of samples, so the file is never read
-  in full. A sample is the 1 MiB at each gigabyte-aligned offset: 0, 1 GiB, 2
-  GiB and so on. The samples are fed into one hash in order. A sample that would
-  run past the end of the file stops there.
-
-The sampled hash is **deliberately probabilistic**. Bytes outside the samples
-and outside the first and last 64 KiB are never read. Two large files that
-differ only in such bytes are reported as duplicates. That is the price of never
-reading a 150 GB file end to end.
-
-Only files of the same size are compared, so their samples always lie at the
-same offsets. A file under 100 MiB and one of 100 MiB or more never share a
-size, so a whole-file hash is never compared with a sampled one.
-
-`scan` reads a file only to hash it, and only when the file is new or has
-changed since its record was written (see "`scan` mode").
 
 ### `scan` mode
 
@@ -466,22 +479,35 @@ FIFO, or device node, or a directory named `.zfs`, is not scanned. `scan` prints
 a one-line warning naming the path and what it is, counts it as skipped, and
 drops it from the scanned operands before reading the database. Another operand
 beneath it is still scanned. The records stored beneath it are not deleted: they
-are treated like any other record outside the scanned operands. If it lies under
+are treated like any other record outside the scanned operands, including the
-another operand, they are under that operand instead, and are treated like any
+content-phase exception below. If it lies under another operand, they are under
-other record there. This is not an error: a scan whose every operand is dropped
+that operand instead, and are treated like any other record there. This is not
-walks nothing and exits 0. An operand that is not on a persistent filesystem
+an error: a scan whose every operand is dropped walks nothing and exits 0. An
-(see "Rules for the walk") is dropped the same way. The line `scan` prints for
+operand that is not on a persistent filesystem (see "Rules for the walk") is
-it is not a warning, though, and it is not counted as skipped.
+dropped the same way, except that the line `scan` prints for it is not a
+warning, and it is not counted as skipped.
 
 `scan` synchronizes the database with the filesystem state under the scanned
 operands:
 
-- A file not yet in the database is hashed and inserted.
+- Only a file whose size at least one other file shares is ever read: a
+  size-unique file cannot be a duplicate, so it is recorded without hashes
+  (`head`, `tail`, and `content` empty). The size census covers every file
+  walked this scan plus every database record outside the scanned operands, so a
+  possible duplicate of a separately scanned tree is still recognized.
+- A file not yet in the database is inserted: hashed when its size is shared,
+  without hashes otherwise.
 - A file already in the database is **skipped without reading its contents**
   when its lstat size equals the recorded size and its lstat mtime is not newer
-  than the recorded mtime. This is what makes a daily rescan cheap.
+  than the recorded mtime. This is what makes a daily rescan cheap. Exception:
-- A file whose mtime is newer than recorded, or whose size differs, is hashed
+  an unchanged file whose record lacks hashes is hashed — and its record updated
-  again. Change detection compares the mtime to the nanosecond, as finely as the
+  — once its size becomes shared, so hashing deferred by size-uniqueness happens
+  as soon as it could matter. Likewise, an unchanged file of 10 MiB or more
+  whose record has no `content` hash is read for one by the content phase below
+  once its size, `head`, and `tail` match another record's.
+- A file whose mtime is newer than recorded, or whose size differs, is processed
+  as if new: re-hashed, or recorded without hashes, per the shared-size rule.
+  Change detection compares the mtime to the nanosecond, as finely as the
   filesystem records it, so a same-size rewrite counts as a change whenever the
   filesystem gives it a later mtime than recorded, even within the same second.
 - A database record whose path lies under one of the scanned operands but was
@@ -489,26 +515,31 @@ operands:
   deleted files. It also removes records for paths that failed to stat or hash
   this run: the database only ever contains signatures verified by the most
   recent scan that covered them (a subsequent successful scan re-adds such
-  files).
+  files). A failure in the content phase below removes nothing: the record is
+  left as it is.
 - Database records outside the scanned operands are untouched, so disjoint trees
-  can be scanned on different schedules into the same database.
+  can be scanned on different schedules into the same database. The one
-- A record at or below a path the walk left out counts as outside the scanned
+  exception is the content phase below: a stored file of 10 MiB or more without
-  operands, so it is not deleted. The walk leaves out a path that:
+  a `content` hash is read for one, wherever it lies, once its size, `head`, and
-    - is a default exclude;
+  `tail` match another record's. If that file is gone, has changed since its
-    - is not on a persistent filesystem;
+  record was written, or is not on a persistent filesystem, it is not read and
-    - with `-x`, is a directory on another filesystem than its operand.
+  the record is left as it is.
+- A record at or below a path the walk left out, because that path is a default
+  exclude, is not on a persistent filesystem, or, with `-x`, is a directory on
+  another filesystem than its operand, counts as outside the scanned operands:
+  it is not deleted, and its size joins the size census. A filesystem mounted
+  under an operand can therefore be scanned on its own schedule into the same
+  database. Records below an operand at or inside such a path are not kept this
+  way, since that operand's own walk verifies them.
 
-    A filesystem mounted under an operand can therefore be scanned on its own
+`scan` runs **five sequential phases over the whole scan**. Parallelism lives
-    schedule into the same database. Records below an operand at or inside such
+inside the walk, hash and content phases; batched database writes begin during
-    a path are not kept this way, since that operand's own walk verifies them.
+the hash phase:
-
-`scan` runs **four sequential phases over the whole scan**. Parallelism lives
-inside the walk and hash phases, and batched database writes begin during the
-hash phase:
 
 1.  **load** — read the existing records. Each record under the operands is kept
-    in memory, by path, with its size and its mtime, for the walk to compare
+    in memory, by path, with its size, its mtime and whether it has hashes, for
-    each file with.
+    the walk to compare each file with; the size of every other record joins the
+    size census.
 2.  **walk + stat** — enumerate the trees under all `PATH` operands concurrently
     with the walk worker pool: every operand seeds the shared queue, and each
     worker reads one directory at a time, handing discovered subdirectories back
@@ -516,23 +547,49 @@ hash phase:
     (while the directory's metadata is still hot). Sequential directory
     enumeration is metadata-latency-bound and takes hours at tens of millions of
     files; per-directory parallelism is what makes the walk tractable on large
-    or busy pools. The walk compares each file with its record as it goes, and
+    or busy pools. The walk builds the size census and resolves unchanged
-    an unchanged file needs nothing more. Every new or changed file is carried
+    already-hashed files on the fly; every other file is carried to the hash
-    to the hash phase as a (path, size, mtime) record.
+    phase as a (path, size, mtime) record.
-3.  **hash** — the worker pool hashes every carried file as described in
+3.  **hash** — with the census complete, each carried file's size decides its
-    "Duplicate detection" above, which gives the file its `head`, `tail` and
+    fate. Size-unique files are never read: new or changed ones are recorded
-    `content`. Zero-length files have constant hashes and are never opened.
+    without hashes in the update phase, unchanged unhashed ones simply keep
-    Files are hashed in **inode order**, which minimizes seeks on spinning
+    their records. Every file with a shared size is hashed by the worker pool as
-    disks. Paths that are hard links to the same inode are **read once**, and
+    described in "Duplicate detection" above: a file under 10 MiB in full, which
-    all share the one result. A hard-link backup farm therefore costs one read
+    gives its `head`, `tail`, and `content` alike, and a larger file only in its
-    per inode, not per path. The phase total counts actual reads, so its
+    end windows, which give its `head` and `tail`; its content hash is left to
-    progress and its time remaining are meaningful. Completed records are
+    the content phase. Zero-length files have constant hashes and are never
-    committed in batched transactions **while hashing runs**. A scan interrupted
+    opened. Files are hashed in **inode order** (minimizing seeks on spinning
-    after hours therefore keeps everything hashed so far, and the next scan
+    disks), and paths that are hard links to the same inode are **read once**,
-    resumes cheaply, skipping records already written.
+    all sharing the one result — a hard-link backup farm costs one read per
-4.  **update** — commit the final partial batch of hashed records. Then delete
+    inode, not per path. The phase total counts actual reads, so progress and
-    the records the scan did not verify: those of vanished files, and of paths
+    ETA are meaningful. Completed records are committed in batched transactions
-    that failed to stat or hash.
+    **while hashing runs**, so a scan interrupted after hours keeps everything
+    hashed so far and the next scan resumes cheaply, skipping records already
+    written.
+4.  **update** — commit the final partial batch, the hash-less records for
+    size-unique new and changed files, and the deletions for records the scan
+    did not verify (vanished files, plus paths that failed to stat or hash).
+5.  **content** — find every record of 10 MiB or more without a `content` hash
+    whose size, `head`, and `tail` equal another record's, anywhere in the
+    database: records from this scan and records stored by earlier scans, inside
+    or outside the scanned operands. SQLite finds them, so only the records to
+    be read are kept in memory, never every file's hashes. Every record sharing
+    their size, `head`, and `tail`, including one that already has a `content`
+    hash, has its file checked with `lstat` first. A file that is gone, is no
+    longer a regular file, has changed (a different size, or an mtime newer than
+    recorded), or is not on a persistent filesystem (see "Rules for the walk")
+    keeps its record as it is and does not count as a match for the others. Any
+    other `lstat` error is warned about and counted as skipped, with the same
+    result. If such a record has no `content` hash, it stays out of duplicate
+    groups; if it has one, it is still reported until a scan covering its own
+    tree updates or removes it. The files that pass and have no `content` hash
+    are read only if at least two of those records pass, so a file whose only
+    matches are stale costs no read; a file that already has a `content` hash is
+    never read again. They are read by a worker pool as in the hash phase, in
+    inode order and once per inode, and their content hashes are committed in
+    batches. A failed read is warned about and counted as skipped; its record
+    keeps an empty `content`, so it is not a duplicate, and a later scan tries
+    again.
 
 Rules for the walk:
 
@@ -555,10 +612,10 @@ Rules for the walk:
 - Filesystem boundaries are crossed by default. With `-x` (long form
   `--one-file-system`, following the GNU `du`/`rsync` convention), never descend
   into a directory on a different filesystem than its `PATH` operand; each
-  operand is bounded by its own filesystem. The records below such a directory
+  operand is bounded by its own filesystem. The records below such a directory,
-  count as outside the scanned operands, as described in "`scan` mode" above:
+  unless they lie below an operand at or inside that directory, count as outside
-  they are not deleted. That does not hold below an operand at or inside that
+  the scanned operands, as described in "`scan` mode" above: they are not
-  directory.
+  deleted, and their sizes join the size census.
 - Never record anything that is not on a persistent filesystem. A filesystem is
   not persistent when the kernel makes it up or keeps it in memory: one of type
   `proc`, `sysfs`, `devtmpfs`, `devpts`, `tmpfs`, `ramfs`, `cgroup`, `cgroup2`,
@@ -575,19 +632,21 @@ Rules for the walk:
   `scan` cannot list them, it fails (exit 1). The walk then never enters a
   directory, and never records a file, on one of those devices, however it
   reaches it: as an operand, below one, or through a bind mount, which keeps the
-  device ID of what it binds. No flag turns this off. For each operand,
+  device ID of what it binds. The content phase never reads a file on one of
-  directory or file it leaves out this way, `scan` prints a line on stderr
+  them either. No flag turns this off. For each operand, directory or file it
-  naming it and the type of its filesystem. It prints that line as it prints a
+  leaves out this way, `scan` prints a line on stderr naming it and the type of
-  warning, so the line never lands inside a progress display:
+  its filesystem, as it prints a warning, so it never lands inside a progress
+  display:
 
     ```
     scan: skipped /dev/shm (tmpfs, not a persistent filesystem)
     ```
 
-    That line is not a warning, and the path is not counted as skipped. An
+    That line is not a warning, and the path is not counted as skipped. Records
-    earlier scan's records at or below it count as outside the scanned operands,
+    that an earlier scan left at or below it, unless they lie below an operand
-    as described in "`scan` mode" above: they are not deleted. That does not
+    at or inside it, count as outside the scanned operands, as described in
-    hold below an operand at or inside it.
+    "`scan` mode" above: they are not deleted, and their sizes join the size
+    census.
 
 - Below an operand, never descend into a default exclude: a path that holds the
   kernel's view of the running system, or temporary files, so that
@@ -619,35 +678,32 @@ Rules for the walk:
 
     That line is not a warning, and the directory is not counted as skipped. A
     default exclude that is not on a persistent filesystem, as `/proc` is, gets
-    the line for its filesystem instead. One that is itself an operand gets no
+    the line for its filesystem instead, and one that is itself an operand gets
-    line, since it is walked as that operand. When the walk does not enter a
+    no line, since it is walked as that operand. The records below a default
-    default exclude, the records below it count as outside the scanned operands,
+    exclude the walk does not enter, unless they lie below an operand at or
-    as described in "`scan` mode" above: they are not deleted. That does not
+    inside it, count as outside the scanned operands, as described in "`scan`
-    hold below an operand at or inside it.
+    mode" above: they are not deleted, and their sizes join the size census.
 
-- On a per-path error, such as permission denied or an unreadable file: print a
+- On any per-path error (permission denied, file vanished between passes,
-  one-line warning to stderr, skip the path, and continue. Per-file errors never
+  unreadable): print a one-line warning to stderr, skip the path, and continue.
-  abort the run; the final summary reports how many were skipped. As specified
+  Per-file errors never abort the run; the final summary reports how many were
-  above, a skipped path that has a database record from an earlier scan loses
+  skipped. As specified above, a skipped path that has a database record from an
-  that record. An unreadable directory subtree likewise loses its records. This
+  earlier scan loses that record, unless it failed only in the content phase, or
-  is accepted: the database mirrors what the latest scan could actually verify.
+  is an operand dropped before the database was read whose record counts as
-  The exception is an operand dropped before the database was read, whose record
+  outside the scanned operands; an unreadable directory subtree likewise loses
-  counts as outside the scanned operands.
+  its records (accepted: the database mirrors what the latest scan could
-- A file or directory removed while the scan runs is not an error. The scan
+  actually verify).
-  prints no line for it and does not count it as skipped. A path gone when the
-  walk reaches it gets no record. A file gone when the hash phase opens it gets
-  none either. The records an earlier scan left at or below such a path are
-  deleted.
 
-Concurrency: the walk phase, which also stats files, and the hash phase each use
+Concurrency: the walk phase (which also stats files), the hash phase, and the
-a worker pool of `--workers` workers (default `runtime.NumCPU()`). The walk
+content phase each use a worker pool of `--workers` workers (default
-parallelizes across directories, and hashing across files. `--workers` must be
+`runtime.NumCPU()`); the walk parallelizes across directories, hashing across
-at least 1: a smaller value is a usage error, reported in one line on stderr
+files. `--workers` must be at least 1: a smaller value is a usage error,
-with exit 2 before anything is scanned. Both phases are seek-bound on spinning
+reported in one line on stderr with exit 2 before anything is scanned. All three
-disks, so raising `--workers` well past the core count can help on pools with
+phases are seek-bound on spinning disks, so raising `--workers` well past the
-many spindles. The main goroutine owns the database writes and the progress
+core count can help on pools with many spindles. The main goroutine owns
-counts. Each progress display draws the counts from a goroutine of its own, and
+partitioning, database writes, and the progress counts, which each progress
-must never block the workers.
+display draws from a goroutine of its own; progress display must never block the
+workers.
 
 `scan` writes nothing to stdout. The summary line on stderr reports the files
 seen this run broken down by disposition, plus skips:
@@ -670,7 +726,9 @@ output whether or not the scanned filesystem is still mounted.
 
 Processing:
 
-- Group the records by the key `(size, head, tail, content)`.
+- Records without a `content` hash (see "Database" above) are excluded: their
+  content is unknown, so they are never reported as duplicates.
+- Group the remaining records by the key `(size, head, tail, content)`.
 - Every group with two or more paths is a duplicate group.
 - Within each group, sort paths lexicographically (byte order). The first path
   is the group's `first`; every other path is a `dupe`.
@@ -724,7 +782,9 @@ records, split on `/`.
 Definitions:
 
 - A file's **signature** is `(size, head, tail, content)` — mtime is
-  informational and excluded.
+  informational and excluded. A record without a `content` hash has unknown
+  content: its signature is treated as unique to that file, so a tree containing
+  such a file never compares equal to any other tree.
 - A directory's **digest** is a SHA-256 Merkle digest computed bottom-up:
   serialize the directory's child entries — for a file child, its name and
   signature; for a subdirectory child, its name and that subdirectory's digest —
@@ -798,6 +858,8 @@ database file size	2147483648 (2.0 GiB)
 modified	2026-10-16T12:02:50Z
 schema version	1
 records	12345678
+records with head and tail hashes	1234567
+records with content hash	123456
 size of recorded files	150000000000000 (136.4 TiB)
 ```
 
@@ -810,8 +872,10 @@ size of recorded files	150000000000000 (136.4 TiB)
   database is in WAL mode, the newest changes may be in the `-wal` file alone.
 - `schema version` is the database's `PRAGMA user_version`. A database with
   another version than this build's is refused, as by `report`.
-- `records` is the number of records. `size of recorded files` adds up the sizes
+- `records` is the number of records. `records with head and tail hashes` and
-  of all records, in bytes, then in human units.
+  `records with content hash` count the records that have those hashes (see
+  "Database"), and `size of recorded files` adds up the sizes of all records, in
+  bytes, then in human units.
 
 `info` runs in two phases, each with its own progress display on stderr (see
 "Progress"): `open` opens the database; `count` reads the schema version and
@@ -827,21 +891,28 @@ Use the progress-bar library for all progress of `scan`, `report`, `trees` and
 stderr.
 
 `scan`, `report`, `trees` and `info` must always show that they are working and
-not hung. Each phase gets its own display from the moment it starts. The display
+not hung. Each phase gets its own display from the moment it starts, naming the
-names the phase and its place among the command's phases. It shows the phase's
+phase and its place among the command's phases and showing its progress, its
-progress, its elapsed time and its time remaining.
+elapsed time and its estimated time remaining (ETA).
 
-A scan has four phases (`[1/4] load`, `[2/4] walk`, `[3/4] hash`,
+A scan has five phases (`[1/5] load`, `[2/5] walk`, `[3/5] hash`,
-`[4/4] update`):
+`[4/5] update`, `[5/5] content`):
 
 - `load` counts the records first, then reads them against that count. While the
   count runs, the display says it is counting the records.
 - `walk` has no exact total. Its estimate is the number of records the last scan
-  left under the operands, known once `load` is done and shown with a `~`. The
+  left under the operands, known once `load` is done and shown with a `~`. Once
-  time remaining is unknown once the walk passes the estimate, and when there
+  the walk passes the estimate, or when there are no such records, as on a first
-  are no such records, as on a first scan.
+  scan, the ETA is unknown.
 - `hash` and `update` have exact totals. Only files that actually need hashing
-  appear in the hash total, so its time remaining is meaningful.
+  appear in the hash total, so its ETA is meaningful.
+- `content` first counts the records whose files it checks with `lstat`, which
+  SQLite does by adding up the number of records with each matching size,
+  `head`, and `tail`, without fetching the records. It then says it is searching
+  the database until SQLite returns the first record; the checks, and their rate
+  and ETA, start there. The files to read then get a display of their own, with
+  their count as its total, and so does the writing of the last of their records
+  to the database, with the count of those records as its total.
 
 `trees` has four phases (`[1/4] open`, `[2/4] read`, `[3/4] group`,
 `[4/4] write`, see "`trees` mode"):
@@ -866,8 +937,8 @@ A scan has four phases (`[1/4] load`, `[2/4] walk`, `[3/4] hash`,
 - `open` has no total. While it opens the database, waiting on a busy one
   included, the display says it is opening the database.
 - `count` has no total either, because one query counts everything: the display
-  says it is counting the records until the query returns, and its time
+  says it is counting the records until the query returns, and its ETA is
-  remaining is unknown.
+  unknown.
 
 Each display is one line holding:
 
@@ -880,69 +951,56 @@ Each display is one line holding:
   while nothing completes, and a burst of items after a wait shows as a burst
   rather than as a steady rate. Once no item has completed for 10 seconds, the
   word `stalled` and where the phase waits take the rate's place (see below).
-- elapsed time, labelled `elapsed`
+- elapsed time
-- time remaining, labelled `remain`: a duration such as `12m9s`, or `unknown`.
+- estimated time remaining, or `unknown`. It comes from the phase's average rate
-  It comes from the phase's average rate since it started counting, so that it
+  since it started counting, so that it does not swing with each burst.
-  does not swing with each burst.
 
-Example lines, as printed when stderr is not a terminal:
+Example shape (exact layout is flexible, content is not):
 
 ```
-[3/4] hash: [12345/98765] 12%, 92 files/s, elapsed 2m32s, remain 17m44s
+[3/5] hash: [12345/98765] 12% |████      | 92 files/s elapsed 2:32 eta 17:54
-[2/4] walk: [1234567/~1200000] 103%, 5120 files/s, elapsed 4m1s, remain unknown
+[2/5] walk: [1234567/~1200000] 103%, 5120 files/s, elapsed 4m1s, eta unknown
-[3/4] hash: [12345/45678] 27%, stalled disk/db, elapsed 4m30s, remain 12m9s
+[3/5] hash: [12345/45678] 27%, stalled disk/db, elapsed 4m30s, eta 12m9s
 ```
 
 When no item has completed for 10 seconds, as while one large file is read, a
 disk is slow or the database is busy, the line says `stalled` and where the
 phase waits in place of the rate, which would read 0: `disk` for files, `db` for
 the database, `out` for stdout, and two of them, as `disk/db`, for a phase that
-may wait on either. `walk` says `disk`. `hash` says `disk/db`, as it reads files
+may wait on either. `walk` says `disk`. `hash` and the content phase's reads say
-and writes their records to the database between them. `load`, `update` and the
+`disk/db`, as they read files and write their records to the database between
-`read` phase of `trees` say `db`. `trees`' `write` says `out`, and its `group`,
+them, and so do the content phase's checks, which `lstat` each record's file as
-which works in memory, says `cpu`. The `read` phase of `report` writes the
+the database search returns it. `load`, `update`, the content phase's last write
-report while it reads the records, so it says `db/out`. A phase still waiting
+and the `read` phase of `trees` say `db`. `trees`' `write` says `out`, and its
-for its total already says what it waits for, and says nothing more.
+`group`, which works in memory, says `cpu`. The `read` phase of `report` writes
+the report while it reads the records, so it says `db/out`. A phase still
+waiting for its total already says what it waits for, and says nothing more.
 
 Additional requirements:
 
-- On a terminal, every display starts with the progress-bar library's spinner in
+- On a terminal, a phase with an exact total (every phase but `scan`'s `walk`,
-  its first column, followed by its line. Each display is drawn again several
+  the `open` phase of `report` and `trees`, and both phases of `info`, once it
-  times a second, so its spinner turns and its elapsed time keeps moving while a
+  has counted what it will do) is drawn as its line followed by the progress-bar
-  phase waits for its next item.
+  library's bar, 10 columns wide, which shows the percent in place of the line's
-- On a terminal, a phase with an exact total ends its line with its percent and
+  own. Any other phase, and a phase whose bar has no room, is drawn as the
-  a bar 10 columns wide. This percent takes the place of the line's own. The bar
+  library's spinner followed by its line. Each display is drawn again several
-  stays at the terminal's right edge, so it does not move as the line grows or
+  times a second, so its elapsed time keeps moving while a phase waits for its
-  shrinks. A phase with nothing to do shows no bar. Every phase has an exact
+  next item.
-  total once it has counted what it will do, except these:
-    - `scan`'s `walk`;
-    - the `open` phase of `report` and `trees`;
-    - both phases of `info`.
-
-    On a terminal 100 columns wide:
-
-    ```
-    ⠹ [3/4] hash: [12345/98765], 92 files/s, elapsed 2m32s, remain 17m44s            12% |█         |
-    ```
-
 - On a terminal no line reaches the terminal's last column, however narrow the
-  terminal is. A line too wide for the terminal first loses whichever of these
+  terminal is. A line too wide for the terminal first loses its bar, which gives
-  it shows: its rate, the words of a stalled phase, or what the phase is waiting
+  way to the spinner, then its rate, the words of a stalled phase, or what the
-  for. The other four parts are kept whole:
+  phase is waiting for. The phase and its place, the progress, the elapsed time
-    - the phase and its place;
+  and the ETA are kept whole. On an 80-column terminal, at counts like those of
-    - the progress;
+  the stalled line in the example above, a stalled line keeps its words whenever
-    - the elapsed time;
+  the elapsed time and the ETA are both in minutes, except in the content
-    - the time remaining.
+  phase's checks and reads. Their lines are the widest, and at most times in
-
+  minutes they lose the words. Where the words and the ETA do not both fit, as
-    The line loses its bar only when those four and the bar do not fit. It is
+  at millions of files and times in hours, the words give way. Only when those
-    cut short at the terminal's edge only when those four alone are too wide,
+  four alone are too wide, as on a terminal narrower than about 70 columns hours
-    and then the end of the time remaining is lost first. That happens hours
+  into a scan of millions of files, is the line cut short at the terminal's
-    into a scan of millions of files, on a terminal narrower than about 75
+  edge, so the end of the ETA is lost first. A terminal made narrower while a
-    columns. On a terminal 100 columns wide, the stalled line of the example
+  phase runs gets the same: the next draw erases the line within the new width
-    above keeps its words and its bar. The next draw fits the line to a terminal
+  and fits the line to it, and the display goes on redrawing in place.
-    made narrower while a phase runs. It erases the old line within the new
-    width, and the display goes on redrawing in place.
-
 - When stderr is not a terminal (a pipe, a file, `/dev/null`), do not emit ANSI
   redraws: print the line the moment each phase starts, every 5 seconds while it
   runs (whether or not an item has completed, and never more often), and once
@@ -1174,8 +1232,8 @@ Tracked in [TODO.md](TODO.md).
 ## Non-goals
 
 - No byte-for-byte compare, and no deletion or linking of duplicates. Files that
-  match are compared by a SHA-256 of the whole file below 100 MiB, and only by
+  match are compared by a SHA-256 of the whole file below 50 MiB, and only by
-  samples at 100 MiB and over. The reports are advisory; acting on them is the
+  samples at 50 MiB and over. The reports are advisory; acting on them is the
   user's job.
 - No persistence beyond the SQLite database described above; no export/import
   formats.

TODO.md

@@ -29,31 +29,6 @@
 
 # Completed Steps
 
-- a file or directory removed while the scan runs is not an error, in the walk
-  or in the hash phase. The scan prints no line for it and does not count it as
-  skipped. When the first of several hard links to a file is gone, the hash
-  phase reads the next one instead. On a terminal every progress display starts
-  with the spinner, the hash phase's included. A phase with an exact total ends
-  its line with its percent and a 10-column bar at the terminal's right edge,
-  which no longer moves as the line's length changes. A line too wide for the
-  terminal now loses its rate before its bar. The time remaining is labelled
-  `remain`, not `eta` (2026-10-26, https://git.eeqj.de/sneak/sfdupes/issues/133)
-
-- the `head`, `tail` and `content` columns are BLOBs, each holding the raw 32
-  bytes of a SHA-256, half the size of the hex text they held. The hashes stay
-  raw from the hash phase through the database to `report` and `trees`. Reading
-  a record whose hash is not 32 bytes is an error (2026-10-26,
-  https://git.eeqj.de/sneak/sfdupes/issues/134)
-
-- the hash phase gives every file the walk records three hashes: `head` and
-  `tail`, the SHA-256 of its first and last 64 KiB, and `content`, the SHA-256
-  of the whole file under 100 MiB and of 1 MiB at each gigabyte from 100 MiB up.
-  The 10 MiB and 50 MiB thresholds, the size census, the records stored without
-  hashes and the content phase are gone, so a scan has four phases and a rescan
-  reads only new and changed files. `info` no longer counts the records with
-  hashes, since every record has them (2026-10-25,
-  https://git.eeqj.de/sneak/sfdupes/issues/132)
-
 - re-vendor `REPO_POLICIES.md` and the four scripts that pass `CACHEBUST` from
   `sneak/prompts` `next` at `0653579`: in the `lint`, `test` and `builder`
   stages of the `Dockerfile` the module download, and in `builder` the

cancel_test.go

@@ -2,7 +2,6 @@ package main
 
 import (
 	"context"
-	"crypto/sha256"
 	"database/sql"
 	"errors"
 	"fmt"
@@ -169,8 +168,8 @@ func assertRecordsIntact(t *testing.T, db *sql.DB, before []string) {
 
 // TestSyncScanCancelledMidWalkKeepsRecords is the regression net under
 // the post-walk guard. The scan is cancelled part-way through the
-// walk, so it reaches the guard having walked a genuinely partial tree
+// walk, so it reaches the guard holding a genuinely partial size
-// and holding a still-populated index of records the walk never got to.
+// census and a still-populated index of records the walk never got to.
 // Every one of those records would look vanished to the update phase.
 // The guard is what stops the scan there, and this test is what
 // notices if it stops doing so: deleting the guard, or making it
@@ -209,10 +208,10 @@ func TestSyncScanCancelledMidWalkKeepsRecords(t *testing.T) {
 }
 
 // assertWalkGuardAborted checks that the scan stopped at the post-walk
-// guard: with a walk that reached some files (it really ran) but not
+// guard: with a census that is neither empty (the walk really ran)
-// all of them (it really was cut short), and with no record counted for
+// nor complete (it really was cut short), and with no record counted
-// removal. A removal count means the partial walk was carried past
+// for removal. A removal count means the partial census was carried
-// the guard into the update phase, which is the failure this test
+// past the guard into the update phase, which is the failure this test
 // exists to catch.
 func assertWalkGuardAborted(t *testing.T, st scanStats, err error) {
 	t.Helper()
@@ -223,30 +222,30 @@ func assertWalkGuardAborted(t *testing.T, st scanStats, err error) {
 	}
 
 	if st.unchanged == 0 {
-		t.Fatalf("stats = %+v: the walk reached no file, so it never "+
+		t.Fatalf("stats = %+v: the census is empty, so the walk never "+
 			"ran and the guard was reached for the wrong reason", st)
 	}
 
 	if st.unchanged >= walkCancelFiles {
-		t.Fatalf("stats = %+v: the walk reached the whole tree, so it "+
+		t.Fatalf("stats = %+v: the census covers the whole tree, so the "+
-			"was not cut short", st)
+			"walk was not cut short", st)
 	}
 
 	// The workers drop every directory still queued once the scan is
 	// cancelled, so only the files in the directories already in flight
-	// can still be reached after the fact. A walk beyond that bound
+	// can add to the census after the fact. A census beyond that bound
 	// would mean the cancellation was not observed where it should have
 	// been.
 	limit := walkCancelAtDone + walkCancelInFlightFiles
 	if st.unchanged > limit {
-		t.Errorf("the walk reached %d files, want at most %d: it kept "+
+		t.Errorf("census covers %d files, want at most %d: the walk kept "+
 			"taking directories off the queue after cancellation",
 			st.unchanged, limit)
 	}
 
 	if st.removed != 0 {
 		t.Errorf("stats = %+v: the scan counted records for removal from "+
-			"a partial walk", st)
+			"a partial census", st)
 	}
 }
 
@@ -763,9 +762,7 @@ func TestHashWorkerDropsQueuedRuns(t *testing.T) {
 
 	var hashed atomic.Bool
 
-	hash := func(path string, size int64) (
+	hash := func(path string, size int64) (string, string, string, error) {
-		[sha256.Size]byte, [sha256.Size]byte, [sha256.Size]byte, error,
-	) {
 		hashed.Store(true)
 
 		return hashSignature(path, size)
@@ -805,9 +802,7 @@ func TestHashWorkerAbandonsBlockedSend(t *testing.T) {
 
 	// The scan is cancelled while the worker hashes, so the worker has
 	// already passed the check that drops queued runs.
-	hash := func(path string, size int64) (
+	hash := func(path string, size int64) (string, string, string, error) {
-		[sha256.Size]byte, [sha256.Size]byte, [sha256.Size]byte, error,
-	) {
 		cancel()
 
 		return hashSignature(path, size)

db.go

@@ -2,7 +2,6 @@ package main
 
 import (
 	"context"
-	"crypto/sha256"
 	"database/sql"
 	"errors"
 	"fmt"
@@ -49,17 +48,16 @@ const lockFilePerm = 0o600
 // createTableSQL is the schema applied to a fresh database. Paths are
 // BLOBs because Unix paths are raw bytes, not guaranteed UTF-8. mtime
 // holds whole Unix seconds and mtime_nsec the nanoseconds within that
-// second. head, tail and content each hold the raw 32 bytes of a
+// second.
-// SHA-256.
 const createTableSQL = `
 CREATE TABLE files (
     path       BLOB PRIMARY KEY,
     size       INTEGER NOT NULL,
     mtime      INTEGER NOT NULL,
     mtime_nsec INTEGER NOT NULL,
-    head       BLOB NOT NULL,
+    head       TEXT NOT NULL,
-    tail       BLOB NOT NULL,
+    tail       TEXT NOT NULL,
-    content    BLOB NOT NULL
+    content    TEXT NOT NULL
 ) WITHOUT ROWID
 `
 
@@ -94,10 +92,6 @@ var errSchemaVersion = errors.New("unsupported database schema version")
 // database.
 var errScanRunning = errors.New("another scan is running")
 
-// errHashSize reports a record whose head, tail or content is not the
-// 32 bytes of a SHA-256.
-var errHashSize = errors.New("hash is not 32 bytes")
-
 // databasePath resolves the database location: SFDUPES_DATABASE when
 // set and non-empty, the compiled-in default otherwise. It returns the
 // path, with the working directory, its symlinks resolved, put in front
@@ -498,24 +492,19 @@ func loadFileRows(ctx context.Context, db *sql.DB, total func(n int64),
 
 	for rows.Next() {
 		var (
-			path, head, tail, content []byte
+			path      []byte
 			sec, nsec int64
 			r         scanRec
 		)
 
-		err = rows.Scan(&path, &r.size, &sec, &nsec, &head, &tail, &content)
+		err = rows.Scan(&path, &r.size, &sec, &nsec, &r.head, &r.tail,
+			&r.content)
 		if err != nil {
 			return fmt.Errorf("read record: %w", err)
 		}
 
 		r.path = string(path)
 		r.mtime = time.Unix(sec, nsec)
-
-		err = r.setHashes(head, tail, content)
-		if err != nil {
-			return fmt.Errorf("read record %s: %w", escapePath(r.path), err)
-		}
-
 		fn(r)
 	}
 
@@ -572,22 +561,17 @@ func loadSignatureRows(ctx context.Context, db *sql.DB,
 
 	for rows.Next() {
 		var (
-			path, head, tail, content []byte
+			path []byte
 			r    scanRec
 		)
 
-		err = rows.Scan(&r.size, &head, &tail, &content, &path)
+		err = rows.Scan(&r.size, &r.head, &r.tail, &r.content, &path)
 		if err != nil {
 			return fmt.Errorf("read record: %w", err)
 		}
 
 		r.path = string(path)
 
-		err = r.setHashes(head, tail, content)
-		if err != nil {
-			return fmt.Errorf("read record %s: %w", escapePath(r.path), err)
-		}
-
 		err = fn(r)
 		if err != nil {
 			return err
@@ -602,22 +586,6 @@ func loadSignatureRows(ctx context.Context, db *sql.DB,
 	return nil
 }
 
-// setHashes sets r's head, tail and content to the values read from
-// those columns. Each must be the raw 32 bytes of a SHA-256.
-func (r *scanRec) setHashes(head, tail, content []byte) error {
-	for _, b := range [][]byte{head, tail, content} {
-		if len(b) != sha256.Size {
-			return fmt.Errorf("%w: found %d bytes", errHashSize, len(b))
-		}
-	}
-
-	r.head = [sha256.Size]byte(head)
-	r.tail = [sha256.Size]byte(tail)
-	r.content = [sha256.Size]byte(content)
-
-	return nil
-}
-
 // countRecords returns the number of records in the database.
 func countRecords(ctx context.Context, db *sql.DB) (int64, error) {
 	var n int64
@@ -630,11 +598,15 @@ func countRecords(ctx context.Context, db *sql.DB) (int64, error) {
 	return n, nil
 }
 
-// databaseCountsSQL counts the records and adds up their sizes, in one
+// databaseCountsSQL counts the records, those with head and tail
+// hashes and those with a content hash, and adds up their sizes, in one
 // pass over the table. A SUM over no rows is NULL, so on an empty table
-// the SUM falls back to 0.
+// each SUM falls back to 0.
 const databaseCountsSQL = `
-SELECT COUNT(*), COALESCE(SUM(size), 0)
+SELECT COUNT(*),
+       COALESCE(SUM(head <> '' AND tail <> ''), 0),
+       COALESCE(SUM(content <> ''), 0),
+       COALESCE(SUM(size), 0)
 FROM files
 `
 
@@ -643,6 +615,8 @@ FROM files
 type databaseCounts struct {
 	schemaVersion int
 	records       int64 // every record
+	headTail      int64 // records with head and tail hashes
+	content       int64 // records with a content hash
 	size          int64 // the sizes of all records added up
 }
 
@@ -670,7 +644,7 @@ func readDatabaseCounts(ctx context.Context,
 	}
 
 	err = tx.QueryRowContext(ctx, databaseCountsSQL).
-		Scan(&c.records, &c.size)
+		Scan(&c.records, &c.headTail, &c.content, &c.size)
 	if err != nil {
 		return c, fmt.Errorf("count records: %w", err)
 	}
@@ -678,15 +652,15 @@ func readDatabaseCounts(ctx context.Context,
 	return c, nil
 }
 
-// loadFileMeta streams every record's path, size and mtime to fn. Scan
+// loadFileMeta streams every record's path, size, mtime, and whether
-// change detection needs no hash values, and skipping the hash columns
+// it carries hashes to fn. Scan change detection needs no hash
-// keeps the scan's in-memory index small on multi-million-file
+// values, and skipping the hash columns keeps the scan's in-memory
-// databases.
+// index small on multi-million-file databases.
 func loadFileMeta(ctx context.Context, db *sql.DB,
-	fn func(path string, size int64, mtime time.Time),
+	fn func(path string, size int64, mtime time.Time, hashed bool),
 ) error {
 	rows, err := db.QueryContext(ctx,
-		"SELECT path, size, mtime, mtime_nsec FROM files")
+		"SELECT path, size, mtime, mtime_nsec, head <> '' FROM files")
 	if err != nil {
 		return fmt.Errorf("read records: %w", err)
 	}
@@ -697,14 +671,94 @@ func loadFileMeta(ctx context.Context, db *sql.DB,
 		var (
 			path            []byte
 			size, sec, nsec int64
+			hashed          int64
 		)
 
-		err = rows.Scan(&path, &size, &sec, &nsec)
+		err = rows.Scan(&path, &size, &sec, &nsec, &hashed)
 		if err != nil {
 			return fmt.Errorf("read record: %w", err)
 		}
 
-		fn(string(path), size, time.Unix(sec, nsec))
+		fn(string(path), size, time.Unix(sec, nsec), hashed != 0)
+	}
+
+	err = rows.Err()
+	if err != nil {
+		return fmt.Errorf("read records: %w", err)
+	}
+
+	return nil
+}
+
+// contentGroupsSQL selects each group (the records sharing a size,
+// head, and tail) of records of at least headTailMin bytes that has more
+// than one record and at least one record without a content hash, with
+// the number of its records.
+const contentGroupsSQL = `
+SELECT size, head, tail, COUNT(*) AS records
+FROM files
+WHERE size >= ? AND head <> ''
+GROUP BY size, head, tail
+HAVING COUNT(*) > 1 AND SUM(content = '') > 0
+`
+
+// contentCandidatesSQL selects every record of the groups
+// contentGroupsSQL selects, with whether each record has a content
+// hash. SQLite does the grouping, so no other record's hashes are
+// loaded into memory; the rows come ordered by size, head, and tail, so
+// each group's rows arrive together.
+const contentCandidatesSQL = `
+SELECT f.path, f.size, f.mtime, f.mtime_nsec, f.head, f.tail, f.content <> ''
+FROM files AS f
+JOIN (` + contentGroupsSQL + `) AS g USING (size, head, tail)
+ORDER BY size, head, tail
+`
+
+// countContentCandidates returns the number of rows of
+// contentCandidatesSQL, adding up the records of the groups
+// contentGroupsSQL selects, so it neither joins nor sorts.
+func countContentCandidates(ctx context.Context, db *sql.DB) (int64, error) {
+	var n int64
+
+	err := db.QueryRowContext(ctx,
+		"SELECT COALESCE(SUM(records), 0) FROM ("+contentGroupsSQL+")",
+		headTailMin).Scan(&n)
+	if err != nil {
+		return 0, fmt.Errorf("count content candidates: %w", err)
+	}
+
+	return n, nil
+}
+
+// loadContentCandidates streams the rows of contentCandidatesSQL to fn:
+// each record, without its content hash, and whether it has one.
+func loadContentCandidates(ctx context.Context, db *sql.DB,
+	fn func(r scanRec, hashed bool),
+) error {
+	rows, err := db.QueryContext(ctx, contentCandidatesSQL, headTailMin)
+	if err != nil {
+		return fmt.Errorf("read records: %w", err)
+	}
+
+	defer func() { _ = rows.Close() }()
+
+	for rows.Next() {
+		var (
+			path      []byte
+			sec, nsec int64
+			r         scanRec
+			hashed    int64
+		)
+
+		err = rows.Scan(&path, &r.size, &sec, &nsec, &r.head, &r.tail,
+			&hashed)
+		if err != nil {
+			return fmt.Errorf("read record: %w", err)
+		}
+
+		r.path = string(path)
+		r.mtime = time.Unix(sec, nsec)
+		fn(r, hashed != 0)
 	}
 
 	err = rows.Err()
@@ -788,8 +842,7 @@ func execUpserts(ctx context.Context, tx *sql.Tx, upserts []scanRec,
 
 	for _, r := range upserts {
 		_, err = st.ExecContext(ctx, []byte(r.path), r.size,
-			r.mtime.Unix(), r.mtime.Nanosecond(),
+			r.mtime.Unix(), r.mtime.Nanosecond(), r.head, r.tail, r.content)
-			r.head[:], r.tail[:], r.content[:])
 		if err != nil {
 			return fmt.Errorf("upsert %s: %w", r.path, err)
 		}

db_test.go

@@ -2,7 +2,6 @@ package main
 
 import (
 	"context"
-	"crypto/sha256"
 	"database/sql"
 	"errors"
 	"fmt"
@@ -474,67 +473,6 @@ func TestLoadRowsCountAgreeDuringScan(t *testing.T) {
 	}
 }
 
-// TestLoadRowsRefuseWrongSizeHash stores a record whose head, tail or
-// content is not the 32 bytes of a SHA-256: either one byte, or the 64
-// bytes of a SHA-256 written as hex text. loadFileRows and
-// loadSignatureRows must each fail on it with an error naming its path.
-func TestLoadRowsRefuseWrongSizeHash(t *testing.T) {
-	t.Parallel()
-
-	hexText := strings.Repeat("ab", sha256.Size)
-
-	cases := []struct {
-		name   string
-		update string
-		value  any
-	}{
-		{"one-byte head", "UPDATE files SET head = ?", []byte{0}},
-		{"hex-text head", "UPDATE files SET head = ?", hexText},
-		{"one-byte tail", "UPDATE files SET tail = ?", []byte{0}},
-		{"hex-text tail", "UPDATE files SET tail = ?", hexText},
-		{"one-byte content", "UPDATE files SET content = ?", []byte{0}},
-		{"hex-text content", "UPDATE files SET content = ?", hexText},
-	}
-	for _, c := range cases {
-		t.Run(c.name, func(t *testing.T) {
-			t.Parallel()
-
-			db := openTestDB(t)
-
-			err := applyChanges(t.Context(), db,
-				[]scanRec{{path: "/a/f"}}, nil, nil)
-			if err != nil {
-				t.Fatal(err)
-			}
-
-			_, err = db.ExecContext(t.Context(), c.update, c.value)
-			if err != nil {
-				t.Fatal(err)
-			}
-
-			loads := map[string]func() error{
-				"loadFileRows": func() error {
-					return loadFileRows(t.Context(), db, func(int64) {},
-						func(scanRec) {})
-				},
-				"loadSignatureRows": func() error {
-					return loadSignatureRows(t.Context(), db, func(int64) {},
-						func(scanRec) error { return nil })
-				},
-			}
-
-			for name, load := range loads {
-				err := load()
-				if !errors.Is(err, errHashSize) ||
-					!strings.Contains(err.Error(), "/a/f") {
-					t.Errorf("%s: err = %v, want %v naming /a/f",
-						name, err, errHashSize)
-				}
-			}
-		})
-	}
-}
-
 // TestSignatureRowsNeedNoSort checks that SQLite reads signatureRowsSQL
 // from the files_signature index alone, with nothing to sort: the
 // first record comes at once, so report's read phase counts each
@@ -589,13 +527,12 @@ func TestApplyChangesRoundTrip(t *testing.T) {
 	// written.
 	recs := []scanRec{
 		{
-			size: 2, mtime: time.Unix(20, 999_999_999), head: testHash("h2"),
+			size: 2, mtime: time.Unix(20, 999_999_999), head: "h2", tail: "t2",
-			tail: testHash("t2"), content: testHash("c2"),
+			content: "c2", path: "/a/tab\tnew\nline",
-			path: "/a/tab\tnew\nline",
 		},
 		{
-			size: 1, mtime: time.Unix(10, 0), head: testHash("h1"),
+			size: 1, mtime: time.Unix(10, 0), head: "h1", tail: "t1",
-			tail: testHash("t1"), content: testHash("c1"), path: "/a/x",
+			content: "c1", path: "/a/x",
 		},
 	}
 
@@ -616,8 +553,8 @@ func TestApplyChangesRoundTrip(t *testing.T) {
 	// An upsert for an existing path updates in place; a delete
 	// removes exactly its path.
 	upd := scanRec{
-		size: 3, mtime: time.Unix(30, 0), head: testHash("h3"),
+		size: 3, mtime: time.Unix(30, 0), head: "h3", tail: "t3", content: "c3",
-		tail: testHash("t3"), content: testHash("c3"), path: "/a/x",
+		path: "/a/x",
 	}
 
 	err = applyChanges(t.Context(), db, []scanRec{upd},
@@ -644,8 +581,8 @@ func TestApplyChangesBatching(t *testing.T) {
 	recs := make([]scanRec, 0, n)
 	for i := range n {
 		recs = append(recs, scanRec{
-			size: int64(i), mtime: time.Unix(1, 0), head: testHash("h"),
+			size: int64(i), mtime: time.Unix(1, 0), head: "h", tail: "t",
-			tail: testHash("t"), path: fmt.Sprintf("/batch/%07d", i),
+			path: fmt.Sprintf("/batch/%07d", i),
 		})
 	}
 
@@ -675,3 +612,62 @@ func TestApplyChangesBatching(t *testing.T) {
 		t.Fatalf("records = %d, want 0", len(got))
 	}
 }
+
+// TestCountContentCandidates checks that countContentCandidates counts
+// the records the content phase's search returns: those of each group
+// of two or more records of headTailMin or more with a head, where at
+// least one has no content hash.
+func TestCountContentCandidates(t *testing.T) {
+	t.Parallel()
+
+	db := openTestDB(t)
+
+	// Each group's records: size, head, and content, one per record.
+	groups := []struct {
+		size     int64
+		head     string
+		contents []string
+	}{
+		{headTailMin, "h", []string{"", "c", "c"}}, // returned
+		{headTailMin + 1, "h", []string{"", ""}},   // returned
+		{headTailMin + 2, "h", []string{"c", "c"}}, // all hashed
+		{headTailMin + 3, "h", []string{""}},       // one record
+		{headTailMin - 1, "h", []string{"", ""}},   // too small
+		{headTailMin + 4, "", []string{"", ""}},    // no head
+	}
+
+	var recs []scanRec
+
+	for _, g := range groups {
+		for _, content := range g.contents {
+			recs = append(recs, scanRec{
+				size: g.size, mtime: time.Unix(1, 0), head: g.head,
+				tail: "t", content: content,
+				path: fmt.Sprintf("/r/%d", len(recs)),
+			})
+		}
+	}
+
+	err := applyChanges(t.Context(), db, recs, nil, nil)
+	if err != nil {
+		t.Fatalf("applyChanges: %v", err)
+	}
+
+	got, err := countContentCandidates(t.Context(), db)
+	if err != nil {
+		t.Fatalf("countContentCandidates: %v", err)
+	}
+
+	var returned int64
+
+	err = loadContentCandidates(t.Context(), db,
+		func(scanRec, bool) { returned++ })
+	if err != nil {
+		t.Fatalf("loadContentCandidates: %v", err)
+	}
+
+	if got != 5 || returned != 5 {
+		t.Errorf("count = %d, records returned = %d, want 5 and 5",
+			got, returned)
+	}
+}

exclude_test.go

@@ -164,8 +164,9 @@ func TestDefaultExcludesHelp(t *testing.T) {
 // TestScanDoesNotEnterDefaultExclude checks that the walk does not
 // enter a default exclude below an operand. It says so on a line of its
 // own and does not count it as skipped. The records below it count as
-// outside the operands: the record of a file gone from it is kept,
+// outside the operands: the record of a file gone from it is kept, and
-// while a file added to it is not recorded.
+// its size joins the size census, while a file added to it is not
+// recorded.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
 func TestScanDoesNotEnterDefaultExclude(t *testing.T) {
@@ -185,6 +186,10 @@ func TestScanDoesNotEnterDefaultExclude(t *testing.T) {
 
 	unseen := writeFile(t, excluded, "new", pattern(2, 10))
 
+	// The same size as gone, so it is hashed only if the census holds the
+	// size of gone's record.
+	added := writeFile(t, dir, "added", pattern(3, 100))
+
 	st, err := syncScan(t.Context(), db, []string{dir}, 1, false,
 		walkRules{excludes: []string{excluded}})
 	if err != nil {
@@ -198,6 +203,11 @@ func TestScanDoesNotEnterDefaultExclude(t *testing.T) {
 	recs := dbRecords(t, db)
 	recordByPath(t, recs, gone)
 
+	if recordByPath(t, recs, added).head == "" {
+		t.Errorf("%s not hashed: the census lacks the size of %s",
+			added, gone)
+	}
+
 	if slices.Contains(recordPaths(recs), unseen) {
 		t.Errorf("%s recorded: the walk entered %s", unseen, excluded)
 	}
@@ -381,8 +391,9 @@ func TestRunScanSkipsNonPersistentOperand(t *testing.T) {
 // TestScanKeepsRecordsOnNonPersistent scans a directory on a filesystem
 // that is not persistent beside one on the tests' temporary directory.
 // The records an earlier scan left in it count as outside the operands:
-// the record of a file gone from it is kept, while a file added to it
+// the record of a file gone from it is kept, and its size joins the size
-// is not recorded.
+// census, while a file added to it is not recorded, and its size does
+// not join the census.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
 func TestScanKeepsRecordsOnNonPersistent(t *testing.T) {
@@ -402,6 +413,11 @@ func TestScanKeepsRecordsOnNonPersistent(t *testing.T) {
 
 	unseen := writeFile(t, shm, "new", pattern(2, 10))
 
+	// added has the size of gone, and lonely that of unseen: each is
+	// hashed only if the census holds that size.
+	added := writeFile(t, dir, "added", pattern(3, 100))
+	lonely := writeFile(t, dir, "lonely", pattern(4, 10))
+
 	st, err := syncScan(t.Context(), db, []string{shm, dir}, 1, false, rules)
 	if err != nil {
 		t.Fatal(err)
@@ -414,6 +430,15 @@ func TestScanKeepsRecordsOnNonPersistent(t *testing.T) {
 	recs := dbRecords(t, db)
 	recordByPath(t, recs, gone)
 
+	if recordByPath(t, recs, added).head == "" {
+		t.Errorf("%s not hashed: the census lacks the size of %s",
+			added, gone)
+	}
+
+	if recordByPath(t, recs, lonely).head != "" {
+		t.Errorf("%s hashed: the census holds the size of %s", lonely, unseen)
+	}
+
 	if slices.Contains(recordPaths(recs), unseen) {
 		t.Errorf("%s recorded", unseen)
 	}
@@ -425,6 +450,51 @@ func TestScanKeepsRecordsOnNonPersistent(t *testing.T) {
 	}
 }
 
+// TestScanContentSkipsNonPersistent checks that the content phase does
+// not read a file on a filesystem that is not persistent, though an
+// earlier scan left a record for it: it is treated as a changed file,
+// so its record is left as it is and its identical copy in another
+// tree, having no other match, is not read either.
+func TestScanContentSkipsNonPersistent(t *testing.T) {
+	t.Parallel()
+
+	shm, _, rules := shmTempDir(t)
+	db := openTestDB(t)
+
+	// A scan that, without the list of filesystems, records a with a
+	// head and tail but no content hash.
+	a := sparseFileWithoutMatch(t, shm, "a", headTailMin)
+
+	syncTree(t, db, shm)
+
+	before := recordByPath(t, dbRecords(t, db), a)
+	if before.head == "" || before.content != "" {
+		t.Fatalf("after scanning %s: %+v, want head and tail only", shm, before)
+	}
+
+	b := sparseFile(t, t.TempDir(), "b", headTailMin)
+
+	st, err := syncScan(t.Context(), db, []string{filepath.Dir(b)}, 1, false,
+		rules)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if st.skipped != 0 {
+		t.Errorf("skipped = %d, want 0", st.skipped)
+	}
+
+	recs := dbRecords(t, db)
+	if r := recordByPath(t, recs, a); r != before {
+		t.Errorf("record = %+v, want it left as %+v", r, before)
+	}
+
+	if r := recordByPath(t, recs, b); r.content != "" {
+		t.Errorf("%s: content = %q, want none: its only match is on %s",
+			b, r.content, shm)
+	}
+}
+
 // TestWalkLeavesOutNonPersistent starts the walk itself in a directory
 // on a filesystem that is not persistent, which a scan never does: it
 // drops such an operand first. Below a directory on a persistent

info.go

@@ -113,6 +113,9 @@ func writeInfo(stdout io.Writer, dbPath, source string,
 		dbFile.ModTime().UTC().Format(time.RFC3339))
 	fmt.Fprintf(&b, "schema version\t%d\n", counts.schemaVersion)
 	fmt.Fprintf(&b, "records\t%d\n", counts.records)
+	fmt.Fprintf(&b, "records with head and tail hashes\t%d\n",
+		counts.headTail)
+	fmt.Fprintf(&b, "records with content hash\t%d\n", counts.content)
 	fmt.Fprintf(&b, "size of recorded files\t%d (%s)\n",
 		counts.size, humanBytes(counts.size))
 

info_test.go

@@ -12,25 +12,23 @@ import (
 // infoOpenLine is the plain line of info's open phase when it opens the
 // database at once; it prints it as the phase starts and as it ends.
 const infoOpenLine = "[1/2] open: opening the database, elapsed 0s, " +
-	"remain unknown"
+	"eta unknown"
 
 // infoCountLine is the plain line of info's count phase when the count
 // takes under a second; it prints it as the phase starts and as it
 // ends.
 const infoCountLine = "[2/2] count: counting the records, elapsed 0s, " +
-	"remain unknown"
+	"eta unknown"
 
-// infoRecs are four records: a duplicate pair, a file of 20 MiB and
+// infoRecs are four records: a duplicate pair with every hash, a file
-// another small file.
+// of 20 MiB with head and tail hashes but no content hash yet, and a
+// file never hashed.
 func infoRecs() []scanRec {
-	h, tl, c, u := testHash("h"), testHash("t"), testHash("c"), testHash("u")
-	bigH, bigT, bigC := testHash("bh"), testHash("bt"), testHash("bc")
-
 	return []scanRec{
-		{size: 5, head: h, tail: tl, content: c, path: "/d/a"},
+		{size: 5, head: "h", tail: "t", content: "c", path: "/d/a"},
-		{size: 5, head: h, tail: tl, content: c, path: "/d/b"},
+		{size: 5, head: "h", tail: "t", content: "c", path: "/d/b"},
-		{size: 20 << 20, head: bigH, tail: bigT, content: bigC, path: "/d/big"},
+		{size: 20 << 20, head: "bh", tail: "bt", path: "/d/big"},
-		{size: 7, head: u, tail: u, content: u, path: "/d/unique"},
+		{size: 7, path: "/d/unique"},
 	}
 }
 
@@ -47,12 +45,16 @@ func TestRunInfo(t *testing.T) {
 		"empty": {
 			want: "schema version\t1\n" +
 				"records\t0\n" +
+				"records with head and tail hashes\t0\n" +
+				"records with content hash\t0\n" +
 				"size of recorded files\t0 (0 B)\n",
 		},
 		"filled": {
 			recs: infoRecs(),
 			want: "schema version\t1\n" +
 				"records\t4\n" +
+				"records with head and tail hashes\t3\n" +
+				"records with content hash\t2\n" +
 				"size of recorded files\t20971537 (20.0 MiB)\n",
 		},
 	}

main.go

@@ -1,12 +1,14 @@
 // Command sfdupes quickly identifies candidate duplicate files across
 // very large filesystems without reading every byte of every file.
 // Files are considered duplicates when their sizes are equal and they
-// agree on three SHA-256 hashes: of the first 64 KiB, of the last 64
+// agree on a short ladder of SHA-256 hashes. A file under 10 MiB is
-// KiB, and of the content. The content hash covers the whole file under
+// hashed in full. A larger file is compared on the hashes of its first
-// 100 MiB, and 1 MiB at each gigabyte at 100 MiB or more, so a large
+// and last 64 KiB, and only when those match another file's is its
-// file is never read in full. scan maintains a persistent SQLite
+// content hash computed and compared: of the whole file when it is
-// database of file signatures (SFDUPES_DATABASE, default
+// under 50 MiB, or of gigabyte-spaced 1 MiB samples when it is 50 MiB
-// /var/lib/sfdupes/db.sqlite) that the other subcommands read.
+// or larger. scan maintains a persistent SQLite database of file
+// signatures (SFDUPES_DATABASE, default /var/lib/sfdupes/db.sqlite)
+// that the other subcommands read.
 //
 // Usage:
 //
@@ -236,7 +238,7 @@ func newScanCommand() *cobra.Command {
 		}),
 	}
 	scanCmd.Flags().IntVar(&scanWorkers, "workers", runtime.NumCPU(),
-		"concurrent workers for the walk and hash phases")
+		"concurrent workers for the walk, hash, and content phases")
 	scanCmd.Flags().BoolVarP(&scanOneFS, "one-file-system", "x", false,
 		"do not cross filesystem boundaries")
 	scanCmd.Flags().BoolVar(&scanNoDefaultExcludes, "no-default-excludes",

progress.go

@@ -9,7 +9,6 @@ import (
 	"strings"
 	"sync"
 	"time"
-	"unicode/utf8"
 
 	"github.com/schollz/progressbar/v3"
 	"golang.org/x/term"
@@ -35,17 +34,10 @@ const stallTime = 10 * time.Second
 // spinnerType selects the progressbar library's spinner style.
 const spinnerType = 14
 
-// barWidth is the width of a phase's bar between its two ends, as in
+// barWidth is the width of the progressbar library's bar between its
-// the example line in README "Progress".
+// two ends, as in the example line in README "Progress".
 const barWidth = 10
 
-// barCell is one column of the filled part of a bar.
-const barCell = "█"
-
-// barColumns is the width of what percentBar returns: the percent in
-// four columns and its sign, a space, and the bar with its two ends.
-const barColumns = barWidth + 8
-
 // percentScale converts a fraction to a percentage.
 const percentScale = 100
 
@@ -53,12 +45,17 @@ const percentScale = 100
 // read, as the progressbar library takes it.
 const fallbackWidth = 80
 
-// spinnerMargin is the columns of a line besides its text: the spinner
+// spinnerMargin is the columns of a spinner's line besides its text:
-// and the three spaces the library draws around the text, and the
+// the spinner and the three spaces the library draws around the text,
-// terminal's last column, which every line stays clear of so that no
+// and the terminal's last column, which every line stays clear of so
-// terminal wraps it.
+// that no terminal wraps it.
 const spinnerMargin = 5
 
+// barMargin is the columns of a bar's line besides its text: the
+// percent and the bar the library draws after the text, the spaces
+// around them, and the terminal's last column.
+const barMargin = barWidth + 11
+
 // countingRecords is what the display of a phase that reads the records
 // says while it counts them, before it has its total.
 const countingRecords = "counting the records"
@@ -88,7 +85,7 @@ func terminalWidth() int {
 	return width
 }
 
-// phaseLabel names phase with its place among phases, as "[2/4] walk".
+// phaseLabel names phase with its place among phases, as "[2/5] walk".
 func phaseLabel(phases []string, phase string) string {
 	return fmt.Sprintf("[%d/%d] %s",
 		slices.Index(phases, phase)+1, len(phases), phase)
@@ -112,13 +109,13 @@ type drawnCount struct {
 }
 
 // progress renders one phase's progress on stderr. On a TTY the
-// progressbar library draws its spinner and then line, fitted to the
+// progressbar library draws line, fitted to the terminal's width: after
-// terminal's width. A phase with an exact total ends the line with its
+// the library's spinner, or, for a phase with an exact total, before the
-// percent and bar, at the terminal's right edge, when the terminal has
+// library's bar when the terminal has room for the bar beside it. It is
-// room for them. It is drawn again every redrawInterval, so its elapsed
+// drawn again every redrawInterval, so its elapsed time keeps moving
-// time keeps moving while the phase waits for its next item. When
+// while the phase waits for its next item. When stderr is not a TTY it
-// stderr is not a TTY it emits no ANSI redraws: it prints line as the
+// emits no ANSI redraws: it prints line as the phase starts, every
-// phase starts, every plainInterval while it runs, and as it ends.
+// plainInterval while it runs, and as it ends.
 //
 // The draws come from a goroutine of their own, so mu guards the fields
 // below it. Every method must be called from the main goroutine. A nil
@@ -134,6 +131,7 @@ type progress struct {
 
 	mu         sync.Mutex
 	bar        *progressbar.ProgressBar // what the library draws; nil until then
+	withBar    bool                     // bar is the library's bar, not its spinner
 	widest     int                      // the widest line bar has drawn, in columns
 	label      string
 	unit       string
@@ -143,7 +141,7 @@ type progress struct {
 	stalledOn  string
 	count      int64
 	start      time.Time // when the phase started, for the elapsed time
-	rateStart  time.Time // when the count started, for the time remaining
+	rateStart  time.Time // when the count started, for the ETA
 	lastItem   time.Time // when the last item was completed, or the count started
 	// drawn holds the count at each draw over the last rateWindow, oldest
 	// first; the rate counts from the first. It keeps the draw before the
@@ -217,34 +215,55 @@ func (p *progress) draw() {
 	}
 
 	if p.width == nil {
-		fmt.Fprintln(os.Stderr, p.line(now, math.MaxInt, false))
+		fmt.Fprintln(os.Stderr, p.line(now, math.MaxInt, true))
 
 		return
 	}
 
 	width := p.width()
-	text := p.line(now, width-spinnerMargin, true)
 
-	// Before each draw the library erases its line with as many spaces as
+	// A phase with an exact total gets the library's bar when its whole
-	// the widest line it has drawn. A display gets a new one when the
+	// line, without the percent the bar shows, fits beside the bar: the
-	// terminal has been made too narrow for that erase, which would wrap
+	// bar gives way before any part of the line. A total of 0 leaves the
-	// and leave a line behind at every draw.
+	// bar nothing to fill.
-	if p.bar == nil || p.widest >= width {
+	text, withBar := "", false
-		p.newBar(width)
+	if p.total > 0 && !p.estimate {
+		text = p.line(now, math.MaxInt, false)
+		withBar = len(text) <= width-barMargin
+	}
+
+	margin := barMargin
+	if !withBar {
+		margin = spinnerMargin
+		text = p.line(now, width-margin, true)
+	}
+
+	// The library makes a bar or a spinner, and before each draw erases
+	// its line with as many spaces as the widest line it has drawn. A
+	// display gets a new one when it turns from one to the other, as its
+	// line grows or the terminal is resized, and when the terminal has
+	// been made too narrow for that erase, which would wrap and leave a
+	// line behind at every draw.
+	if p.bar == nil || p.withBar != withBar || p.widest >= width {
+		p.newBar(withBar, width)
+	}
+
+	if withBar {
+		_ = p.bar.Set64(p.count)
 	}
 
 	p.bar.Describe(text)
 
-	// The library drew the spinner, the text and the spaces around it,
+	// The library drew the text and its margin, less the terminal's last
-	// less the terminal's last column, which spinnerMargin keeps clear. A
+	// column, which the margin keeps clear.
-	// bar's cells take one column each but more than one byte.
+	p.widest = max(p.widest, len(text)+margin-1)
-	p.widest = max(p.widest, utf8.RuneCountInString(text)+spinnerMargin-1)
 }
 
-// newBar has the library make what draw draws: its spinner, followed by
+// newBar has the library make what draw draws: a bar for the phase's
-// the text draw gives it and nothing of the library's own. p.mu must be
+// total when withBar is set, and a spinner when not. Either shows the
-// held.
+// text draw gives it, and nothing of its own besides the spinner, or
-func (p *progress) newBar(width int) {
+// the bar and its percent. p.mu must be held.
+func (p *progress) newBar(withBar bool, width int) {
 	if p.bar != nil {
 		// Erase what was drawn before, short of the last column of a
 		// terminal that may have been made narrower since: the library's
@@ -253,36 +272,29 @@ func (p *progress) newBar(width int) {
 		fmt.Fprint(os.Stderr, "\r"+blank+"\r")
 	}
 
-	// A total of -1 gives the library's spinner.
+	total := int64(-1) // the library's spinner
-	p.bar = progressbar.NewOptions64(-1,
+	if withBar {
+		total = p.total
+	}
+
+	p.bar = progressbar.NewOptions64(total,
 		progressbar.OptionSetWriter(os.Stderr),
+		progressbar.OptionSetWidth(barWidth),
 		progressbar.OptionSpinnerType(spinnerType),
 		// The spinner turns at each draw, not on a timer of the
 		// library's own.
 		progressbar.OptionSetSpinnerChangeInterval(0),
-		// The text carries the elapsed time and the time remaining.
+		// The text carries the elapsed time and the ETA.
 		progressbar.OptionSetElapsedTime(false),
 		progressbar.OptionSetPredictTime(false),
 	)
-	p.widest = 0
+	p.withBar, p.widest = withBar, 0
-}
-
-// percentBar returns the percent and the bar that end the text of a
-// phase with an exact total, as "  25% |██        |": the bar's
-// barWidth columns fill in proportion to the items done. A count past
-// the total, as when a scan adds records while trees reads them, shows
-// as 100%. p.mu must be held.
-func (p *progress) percentBar() string {
-	done := min(float64(p.count)/float64(p.total), 1)
-	filled := strings.Repeat(barCell, int(done*barWidth))
-
-	return fmt.Sprintf("%4.0f%% |%-*s|", done*percentScale, barWidth, filled)
 }
 
 // setTotal gives a phase that started without a total its total, once
-// it is known. The rate, the time remaining and the time with no item
+// it is known. The rate, the ETA and the time with no item completed
-// completed count from then on: the time spent waiting for the total
+// count from then on: the time spent waiting for the total says nothing
-// says nothing about how fast items go.
+// about how fast items go.
 func (p *progress) setTotal(total int64) {
 	if p == nil {
 		return
@@ -426,26 +438,22 @@ func (p *progress) finish() {
 // line formats the phase's progress as of now in at most room columns,
 // as
 //
-//	[3/4] hash: [1234/5000] 25%, 300 files/s, elapsed 4s, remain 13s
+//	[3/5] hash: [1234/5000] 25%, 300 files/s, elapsed 4s, eta 13s
 //
-// where "files" is the phase's unit. A phase with no total shows the
+// where "files" is the phase's unit, without the percent when percent
-// items done instead, or what it waits for. The rate is that since the
+// is false, as before the library's bar, which shows its own. A phase
-// first draw in drawn, and the time remaining comes from the rate since
+// with no total shows the items done instead, or what it waits for. The
-// the count started. Once no item has been completed for stallTime, the
+// rate is that since the first draw in drawn, and the ETA comes from
-// rate, which would read 0, gives way to the word "stalled" and where
+// the rate since the count started. Once no item has been completed for
-// the phase waits, as "stalled disk/db".
+// stallTime, the rate, which would read 0, gives way to the word
+// "stalled" and where the phase waits, as "stalled disk/db".
 //
-// When bar is true, as on a terminal, a phase with an exact total ends
+// The phase and its place, the progress, the elapsed time and the ETA
-// the line with its percent and bar, in place of the percent after its
+// are kept whole. A line wider than room leaves out the rate, the words
-// count. A total of 0 leaves the bar nothing to fill.
+// of a stalled phase, or what the phase waits for; only if it is still
-//
+// wider is it cut at room. Every part is ASCII, so its length is its
-// The phase and its place, the progress, the elapsed time and the time
+// width in columns. p.mu must be held.
-// remaining are kept whole. A line wider than room first leaves out the
+func (p *progress) line(now time.Time, room int, percent bool) string {
-// rate, the words of a stalled phase, or what the phase waits for. Next
-// it leaves out the bar. Only if it is still wider is it cut at room.
-// Every part but the bar is ASCII, so its length is its width in
-// columns. p.mu must be held.
-func (p *progress) line(now time.Time, room int, bar bool) string {
 	elapsed := now.Sub(p.start).Round(time.Second)
 
 	// head and tail hold the parts kept whole; extra, between them, is
@@ -454,41 +462,21 @@ func (p *progress) line(now time.Time, room int, bar bool) string {
 
 	if p.total < 0 && p.waitingFor != "" {
 		head, extra = p.label+": ", p.waitingFor+", "
-		tail = fmt.Sprintf("elapsed %s, remain unknown", elapsed)
+		tail = fmt.Sprintf("elapsed %s, eta unknown", elapsed)
 	} else {
 		since := p.drawn[0]
 		rate := perSecond(p.count-since.count, now.Sub(since.at))
 		average := perSecond(p.count, now.Sub(p.rateStart))
 
-		head = p.label + ": " + p.itemsDone(true) + ", "
+		head = p.label + ": " + p.itemsDone(percent) + ", "
 		extra = fmt.Sprintf("%.0f %s/s, ", rate, p.unit)
-		tail = fmt.Sprintf("elapsed %s, remain %s", elapsed, p.remain(average))
+		tail = fmt.Sprintf("elapsed %s, eta %s", elapsed, p.eta(average))
 
 		if now.Sub(p.lastItem) >= stallTime {
 			extra = "stalled " + p.stalledOn + ", "
 		}
 	}
 
-	if bar && p.total > 0 && !p.estimate {
-		// Beside the bar, which shows the percent, the count shows none.
-		// The line is padded to the room the bar leaves it, so the bar
-		// stays at room's right edge however long the line is.
-		beside := room - barColumns
-		barHead := p.label + ": " + p.itemsDone(false) + ", "
-
-		switch {
-		case len(barHead+extra+tail) <= beside:
-			return fmt.Sprintf("%-*s%s", beside, barHead+extra+tail,
-				p.percentBar())
-		case len(barHead+tail) <= beside:
-			return fmt.Sprintf("%-*s%s", beside, barHead+tail, p.percentBar())
-		}
-
-		// The bar gives way. The rate, or the words of a stalled phase,
-		// gave way before it and stay out.
-		extra = ""
-	}
-
 	full, kept := head+extra+tail, head+tail
 
 	switch {
@@ -538,10 +526,10 @@ func (p *progress) itemsDone(percent bool) string {
 	return fmt.Sprintf("%s %.0f%%", done, pct)
 }
 
-// remain returns the time the phase has left at rate items a second, or
+// eta returns the time the phase has left at rate items a second, or
 // "unknown": with no total, before the first item, and once the count
 // passes an estimate. p.mu must be held.
-func (p *progress) remain(rate float64) string {
+func (p *progress) eta(rate float64) string {
 	switch {
 	case p.total < 0, p.estimate && p.count > p.total:
 		return "unknown"
@@ -551,8 +539,8 @@ func (p *progress) remain(rate float64) string {
 		return "unknown"
 	}
 
-	left := time.Duration(
+	remaining := time.Duration(
 		float64(p.total-p.count) / rate * float64(time.Second))
 
-	return left.Round(time.Second).String()
+	return remaining.Round(time.Second).String()
 }

progress_test.go

@@ -23,7 +23,7 @@ const testRedraw = time.Millisecond
 const testDeadline = 5 * time.Second
 
 // hashStalledOn is where the hash phase says it waits once it has
-// stalled, as hashPhase gives it.
+// stalled, as readRuns gives it.
 const hashStalledOn = "disk/db"
 
 // testClock is a clock a test moves by hand, for a display that reads
@@ -117,38 +117,38 @@ func TestProgressLine(t *testing.T) {
 			label: scanPhase("load"), total: -1,
 			waitingFor: countingRecords,
 		},
-		want: "[1/4] load: counting the records, elapsed 10s, remain unknown",
+		want: "[1/5] load: counting the records, elapsed 10s, eta unknown",
 	}, {
 		p: &progress{
 			label: scanPhase("hash"), unit: scanUnit, total: 100, count: 20,
 		},
-		want: "[3/4] hash: [20/100] 20%, 2 files/s, elapsed 10s, remain 40s",
+		want: "[3/5] hash: [20/100] 20%, 2 files/s, elapsed 10s, eta 40s",
 	}, {
 		p:    &progress{label: scanPhase("hash"), unit: scanUnit, total: 100},
-		want: "[3/4] hash: [0/100] 0%, 0 files/s, elapsed 10s, remain unknown",
+		want: "[3/5] hash: [0/100] 0%, 0 files/s, elapsed 10s, eta unknown",
 	}, {
 		p:    &progress{label: scanPhase("update"), unit: scanUnit, total: 0},
-		want: "[4/4] update: [0/0] 100%, 0 files/s, elapsed 10s, remain 0s",
+		want: "[4/5] update: [0/0] 100%, 0 files/s, elapsed 10s, eta 0s",
 	}, {
 		p: &progress{
 			label: scanPhase("walk"), unit: scanUnit, total: 100,
 			estimate: true, count: 20,
 		},
-		want: "[2/4] walk: [20/~100] 20%, 2 files/s, elapsed 10s, remain 40s",
+		want: "[2/5] walk: [20/~100] 20%, 2 files/s, elapsed 10s, eta 40s",
 	}, {
 		// Past the estimate the walk cannot tell how long it has left.
 		p: &progress{
 			label: scanPhase("walk"), unit: scanUnit, total: 100,
 			estimate: true, count: 120,
 		},
-		want: "[2/4] walk: [120/~100] 120%, 12 files/s, elapsed 10s, " +
+		want: "[2/5] walk: [120/~100] 120%, 12 files/s, elapsed 10s, " +
-			"remain unknown",
+			"eta unknown",
 	}, {
 		// A first scan's walk has no estimate.
 		p: &progress{
 			label: scanPhase("walk"), unit: scanUnit, total: -1, count: 20,
 		},
-		want: "[2/4] walk: 20 files, 2 files/s, elapsed 10s, remain unknown",
+		want: "[2/5] walk: 20 files, 2 files/s, elapsed 10s, eta unknown",
 	}, {
 		// The rate is in what the phase counts.
 		p: &progress{
@@ -156,14 +156,14 @@ func TestProgressLine(t *testing.T) {
 			count: 20,
 		},
 		want: "[3/4] group: [20/100] 20%, 2 directories/s, elapsed 10s, " +
-			"remain 40s",
+			"eta 40s",
 	}}
 
 	for _, tc := range cases {
 		tc.p.start, tc.p.rateStart, tc.p.lastItem = ago, ago, now
 		tc.p.drawn = []drawnCount{{at: ago}}
 
-		if got := tc.p.line(now, math.MaxInt, false); got != tc.want {
+		if got := tc.p.line(now, math.MaxInt, true); got != tc.want {
 			t.Errorf("line = %q, want %q", got, tc.want)
 		}
 	}
@@ -184,9 +184,9 @@ func TestLineFitsRoom(t *testing.T) {
 		start: ago, rateStart: ago, lastItem: now,
 		drawn: []drawnCount{{at: ago}},
 	}
-	load := &progress{
+	content := &progress{
-		label: scanPhase("load"), total: -1,
+		label: scanPhase("content"), total: -1,
-		waitingFor: countingRecords, start: ago, rateStart: ago,
+		waitingFor: "searching the database", start: ago, rateStart: ago,
 	}
 
 	started := now.Add(-50 * time.Second)
@@ -200,47 +200,46 @@ func TestLineFitsRoom(t *testing.T) {
 	cases := []struct {
 		p       *progress
 		room    int
-		bar  bool
+		percent bool
 		want    string
 	}{{
-		p: hash, room: 60,
+		p: hash, room: 57, percent: true,
-		want: "[3/4] hash: [20/100] 20%, 2 files/s, elapsed 10s, remain 40s",
+		want: "[3/5] hash: [20/100] 20%, 2 files/s, elapsed 10s, eta 40s",
 	}, {
-		p: hash, room: 59,
+		p: hash, room: 56, percent: true,
-		want: "[3/4] hash: [20/100] 20%, elapsed 10s, remain 40s",
+		want: "[3/5] hash: [20/100] 20%, elapsed 10s, eta 40s",
 	}, {
-		p: hash, room: 48,
+		p: hash, room: 45, percent: true,
-		want: "[3/4] hash: [20/100] 20%, elapsed 10s, remain 40",
+		want: "[3/5] hash: [20/100] 20%, elapsed 10s, eta 40",
 	}, {
-		// The bar shows the percent.
+		// Before the library's bar, which shows the percent.
-		p: hash, room: 74, bar: true,
+		p: hash, room: 57, percent: false,
-		want: "[3/4] hash: [20/100], 2 files/s, elapsed 10s, remain 40s" +
+		want: "[3/5] hash: [20/100], 2 files/s, elapsed 10s, eta 40s",
-			"  20% |██        |",
 	}, {
-		p: load, room: 43,
+		p: content, room: 40, percent: true,
-		want: "[1/4] load: elapsed 10s, remain unknown",
+		want: "[5/5] content: elapsed 10s, eta unknown",
 	}, {
-		p: stalled, room: 68,
+		p: stalled, room: 65, percent: true,
-		want: "[3/4] hash: [20/100] 20%, stalled disk/db, elapsed 50s, " +
+		want: "[3/5] hash: [20/100] 20%, stalled disk/db, elapsed 50s, " +
-			"remain 3m20s",
+			"eta 3m20s",
 	}, {
-		p: stalled, room: 67,
+		p: stalled, room: 64, percent: true,
-		want: "[3/4] hash: [20/100] 20%, elapsed 50s, remain 3m20s",
+		want: "[3/5] hash: [20/100] 20%, elapsed 50s, eta 3m20s",
 	}}
 
 	for _, tc := range cases {
-		if got := tc.p.line(now, tc.room, tc.bar); got != tc.want {
+		if got := tc.p.line(now, tc.room, tc.percent); got != tc.want {
-			t.Errorf("line(%d, %t) = %q, want %q", tc.room, tc.bar, got,
+			t.Errorf("line(%d, %t) = %q, want %q",
-				tc.want)
+				tc.room, tc.percent, got, tc.want)
 		}
 	}
 }
 
 // TestSetTotalStartsRate checks that setTotal starts a phase's rate,
-// its time remaining and its time with no record read from when the
+// its ETA and its time with no record read from when the total becomes
-// total becomes known, not from when the phase started waiting for it:
+// known, not from when the phase started waiting for it: after 20
-// after 20 seconds of waiting for the total, the phase has not stalled 5
+// seconds of waiting for the total, the phase has not stalled 5 seconds
-// seconds after it, and has 10 seconds after it with no record read.
+// after it, and has 10 seconds after it with no record read.
 func TestSetTotalStartsRate(t *testing.T) {
 	t.Parallel()
 
@@ -256,15 +255,15 @@ func TestSetTotalStartsRate(t *testing.T) {
 
 	clock.Add(5 * time.Second)
 
-	want := "[1/4] load: [0/100] 0%, 0 files/s, elapsed 25s, remain unknown"
+	want := "[1/5] load: [0/100] 0%, 0 files/s, elapsed 25s, eta unknown"
-	if got := p.line(clock.Now(), math.MaxInt, false); got != want {
+	if got := p.line(clock.Now(), math.MaxInt, true); got != want {
 		t.Errorf("line = %q, want %q", got, want)
 	}
 
 	clock.Add(5 * time.Second)
 
-	want = "[1/4] load: [0/100] 0%, stalled db, elapsed 30s, remain unknown"
+	want = "[1/5] load: [0/100] 0%, stalled db, elapsed 30s, eta unknown"
-	if got := p.line(clock.Now(), math.MaxInt, false); got != want {
+	if got := p.line(clock.Now(), math.MaxInt, true); got != want {
 		t.Errorf("line = %q, want %q", got, want)
 	}
 
@@ -273,8 +272,8 @@ func TestSetTotalStartsRate(t *testing.T) {
 		p.increment()
 	}
 
-	want = "[1/4] load: [20/100] 20%, 2 files/s, elapsed 30s, remain 40s"
+	want = "[1/5] load: [20/100] 20%, 2 files/s, elapsed 30s, eta 40s"
-	if got := p.line(clock.Now(), math.MaxInt, false); got != want {
+	if got := p.line(clock.Now(), math.MaxInt, true); got != want {
 		t.Errorf("line = %q, want %q", got, want)
 	}
 }
@@ -317,9 +316,9 @@ func TestPlainProgressPrintsAtStartAndEnd(t *testing.T) {
 
 	lines := strings.Split(strings.TrimSuffix(stderr(), "\n"), "\n")
 
-	want := "[3/4] hash: [0/1000] 0%, 0 files/s, elapsed 0s, remain unknown"
+	want := "[3/5] hash: [0/1000] 0%, 0 files/s, elapsed 0s, eta unknown"
 	if len(lines) != 2 || lines[0] != want ||
-		!strings.HasPrefix(lines[1], "[3/4] hash: [1000/1000] 100%, ") {
+		!strings.HasPrefix(lines[1], "[3/5] hash: [1000/1000] 100%, ") {
 		t.Errorf("stderr lines = %q, want %q, then the line at 1000",
 			lines, want)
 	}
@@ -337,7 +336,7 @@ func TestPlainProgressPrintsWhileWaiting(t *testing.T) {
 	}, nil, testRedraw, time.Now)
 
 	awaitOutput(t, stderr,
-		"[1/4] load: counting the records, elapsed 0s, remain unknown\n", 3)
+		"[1/5] load: counting the records, elapsed 0s, eta unknown\n", 3)
 	p.finish()
 }
 
@@ -367,23 +366,23 @@ func TestPlainProgressStalledProducer(t *testing.T) {
 
 	clock.Add(5 * time.Second)
 	awaitOutput(t, stderr,
-		"[3/4] hash: [20/100] 20%, 4 files/s, elapsed 5s, remain 20s\n", 3)
+		"[3/5] hash: [20/100] 20%, 4 files/s, elapsed 5s, eta 20s\n", 3)
 
 	clock.Add(time.Second)
 	awaitOutput(t, stderr,
-		"[3/4] hash: [20/100] 20%, 0 files/s, elapsed 6s, remain 24s\n", 3)
+		"[3/5] hash: [20/100] 20%, 0 files/s, elapsed 6s, eta 24s\n", 3)
 
 	clock.Add(4 * time.Second)
 	awaitOutput(t, stderr,
-		"[3/4] hash: [20/100] 20%, stalled disk/db, elapsed 10s, remain 40s\n",
+		"[3/5] hash: [20/100] 20%, stalled disk/db, elapsed 10s, eta 40s\n",
 		3)
 }
 
 // TestPlainProgressRateAfterStall draws a phase's plain line by hand, a
 // little over plainInterval after the line before, as a late tick
 // does, through a stall and a burst of files after it: the rate covers
-// the files read since the line before, while the time remaining comes
+// the files read since the line before, while the ETA comes from the
-// from the phase's average.
+// phase's average.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
 func TestPlainProgressRateAfterStall(t *testing.T) {
@@ -411,10 +410,10 @@ func TestPlainProgressRateAfterStall(t *testing.T) {
 	p.finish()
 
 	want := []string{
-		"[3/4] hash: [0/1000] 0%, 0 files/s, elapsed 0s, remain unknown",
+		"[3/5] hash: [0/1000] 0%, 0 files/s, elapsed 0s, eta unknown",
-		"[3/4] hash: [0/1000] 0%, 0 files/s, elapsed 5s, remain unknown",
+		"[3/5] hash: [0/1000] 0%, 0 files/s, elapsed 5s, eta unknown",
-		"[3/4] hash: [0/1000] 0%, stalled disk/db, elapsed 10s, remain unknown",
+		"[3/5] hash: [0/1000] 0%, stalled disk/db, elapsed 10s, eta unknown",
-		"[3/4] hash: [100/1000] 10%, 20 files/s, elapsed 15s, remain 2m15s",
+		"[3/5] hash: [100/1000] 10%, 20 files/s, elapsed 15s, eta 2m15s",
 	}
 
 	lines := strings.Split(strings.TrimSuffix(stderr(), "\n"), "\n")
@@ -427,8 +426,8 @@ func TestPlainProgressRateAfterStall(t *testing.T) {
 // redrawInterval, as a terminal does, through a stall of stallTime and
 // a burst of files after it: the rate covers the files read over the
 // last rateWindow, and falls to 0 once the burst is older than that,
-// while the time remaining comes from the phase's average. Plain lines
+// while the ETA comes from the phase's average. Plain lines show each
-// show each draw.
+// draw.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
 func TestProgressRateAfterStallRedrawn(t *testing.T) {
@@ -458,8 +457,8 @@ func TestProgressRateAfterStallRedrawn(t *testing.T) {
 	redraw(rateWindow + redrawInterval)
 	p.finish()
 
-	want := "[3/4] hash: [50/1000] 5%, 10 files/s, elapsed 15s, remain 4m45s\n" +
+	want := "[3/5] hash: [50/1000] 5%, 10 files/s, elapsed 15s, eta 4m45s\n" +
-		"[3/4] hash: [50/1000] 5%, 0 files/s, elapsed 15s, remain 4m47s\n"
+		"[3/5] hash: [50/1000] 5%, 0 files/s, elapsed 15s, eta 4m47s\n"
 	if !strings.Contains(stderr(), want) {
 		t.Errorf("stderr = %q, want the lines %q", stderr(), want)
 	}
@@ -481,10 +480,10 @@ func lastFrame(out string) string {
 }
 
 // TestTerminalRedrawsWhileWaiting checks that the terminal display of a
-// phase, with a bar or without, keeps being drawn while the phase waits
+// phase, spinner or bar, keeps being drawn while the phase waits for its
-// for its next item, with the elapsed time as of now, and shows every
+// next item, with the elapsed time as of now, and shows every item
-// item completed before the wait. The terminal is 120 columns wide, so
+// completed before the wait. stderr is not a terminal here, so the
-// the hash phase's display has room for its bar.
+// displays take it to be fallbackWidth wide.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
 func TestTerminalRedrawsWhileWaiting(t *testing.T) {
@@ -507,8 +506,7 @@ func TestTerminalRedrawsWhileWaiting(t *testing.T) {
 	for _, tc := range cases {
 		before := len(stderr())
 		since := func() string { return stderr()[before:] }
-		p := startProgress(tc.params, func() int { return 120 }, testRedraw,
+		p := startProgress(tc.params, terminalWidth, testRedraw, time.Now)
-			time.Now)
 
 		for range 50 {
 			p.increment()
@@ -554,13 +552,13 @@ func drawLate(stderr func() string, params progressParams, width int,
 	return stderr()[before:]
 }
 
-// TestTerminalLinesKeepTimeRemaining checks the line each display shows
+// TestTerminalLinesKeepETA checks the line each display shows on an
-// on an 80-column terminal at a large scan's counts and times, and the
+// 80-column terminal at a large scan's counts and times, and the line
-// line of trees' read phase at the same counts: the rate is left out,
+// of trees' read phase at the same counts: the rate is left out, so
-// so that the elapsed time and the time remaining stay whole.
+// that the elapsed time and the ETA stay whole.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
-func TestTerminalLinesKeepTimeRemaining(t *testing.T) {
+func TestTerminalLinesKeepETA(t *testing.T) {
 	stderr := captureStderr(t)
 
 	cases := []struct {
@@ -572,20 +570,26 @@ func TestTerminalLinesKeepTimeRemaining(t *testing.T) {
 			label: scanPhase("walk"), unit: scanUnit, total: 1200000,
 			estimate: true,
 		},
-		want: "[2/4] walk: [1234567/~1200000] 103%, elapsed 2h30m0s, " +
+		want: "[2/5] walk: [1234567/~1200000] 103%, elapsed 2h30m0s, " +
-			"remain unknown",
+			"eta unknown",
 	}, {
 		params: progressParams{
 			label: scanPhase("hash"), unit: scanUnit, total: 9876543,
 		},
-		want: "[3/4] hash: [1234567/9876543] 12%, elapsed 2h30m0s, " +
+		want: "[3/5] hash: [1234567/9876543] 12%, elapsed 2h30m0s, " +
-			"remain 17h30m0s",
+			"eta 17h30m0s",
+	}, {
+		params: progressParams{
+			label: scanPhase("content"), unit: scanUnit, total: 9876543,
+		},
+		want: "[5/5] content: [1234567/9876543] 12%, elapsed 2h30m0s, " +
+			"eta 17h30m0s",
 	}, {
 		params: progressParams{
 			label: treesPhase("read"), unit: "records", total: 9876543,
 		},
 		want: "[2/4] read: [1234567/9876543] 12%, elapsed 2h30m0s, " +
-			"remain 17h30m0s",
+			"eta 17h30m0s",
 	}}
 
 	for _, tc := range cases {
@@ -596,32 +600,30 @@ func TestTerminalLinesKeepTimeRemaining(t *testing.T) {
 	}
 }
 
-// TestTerminalStalledLines draws every phase's line once the phase has
+// TestTerminalStalledLines draws every phase's line on an 80-column
-// stalled, its last item done 42 seconds before. At an ordinary scan's
+// terminal once the phase has stalled, its last item done 42 seconds
-// counts and times, 12345 of 45678 items done four and a half minutes
+// before. At an ordinary scan's counts and times, 12345 of 45678 items
-// in, on a terminal 100 columns wide, the line shows where the phase
+// done four and a half minutes in, the line shows where the phase waits
-// waits whole, beside the phase, progress, elapsed time and time
+// whole, beside the phase, progress, elapsed time and ETA. At a large
-// remaining. At a large scan's, 1234567 of 9876543 two and a half hours
+// scan's, 1234567 of 9876543 two and a half hours in, those four still
-// in, on an 80-column terminal, those four still show whole, as the
+// show whole, as the words give way first where they do not fit.
-// words give way first where they do not fit.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
 func TestTerminalStalledLines(t *testing.T) {
 	stderr := captureStderr(t)
 
-	// drawStalled draws the display params describes on a terminal width
+	// drawStalled draws the display params describes as it starts and
-	// columns wide as it starts and once more as it ends, with count of
+	// once more as it ends, with count of total items done elapsed into
-	// total items done elapsed into the phase, and returns what the
+	// the phase, and returns what the terminal shows.
-	// terminal shows.
+	drawStalled := func(params progressParams, total, count int64,
-	drawStalled := func(params progressParams, width int, total,
+		elapsed time.Duration,
-		count int64, elapsed time.Duration,
 	) string {
 		before := len(stderr())
 		clock := &testClock{now: time.Now()}
 		params.total = total
 
 		// No timed draws come between the two.
-		p := startProgress(params, func() int { return width }, time.Hour,
+		p := startProgress(params, func() int { return 80 }, time.Hour,
 			clock.Now)
 
 		clock.Add(elapsed - 42*time.Second)
@@ -641,6 +643,9 @@ func TestTerminalStalledLines(t *testing.T) {
 		{label: scanPhase("walk"), estimate: true, stalledOn: "disk"},
 		{label: scanPhase("hash"), stalledOn: hashStalledOn},
 		{label: scanPhase("update"), stalledOn: "db"},
+		// The content phase's checks and reads, then its last write.
+		{label: scanPhase("content"), stalledOn: "disk/db"},
+		{label: scanPhase("content"), stalledOn: "db"},
 		{label: treesPhase("read"), stalledOn: "db"},
 		{label: treesPhase("group"), stalledOn: "cpu"},
 		{label: treesPhase("write"), stalledOn: "out"},
@@ -653,22 +658,19 @@ func TestTerminalStalledLines(t *testing.T) {
 			approx = "~"
 		}
 
-		// The percent is in the line, or beside the bar.
+		shown := drawStalled(params, 45678, 12345,
-		shown := drawStalled(params, 100, 45678, 12345,
 			4*time.Minute+30*time.Second)
-		for _, want := range []string{
+		want := params.label + ": [12345/" + approx + "45678] 27%, stalled " +
-			params.label + ": [12345/" + approx + "45678]",
+			params.stalledOn + ", elapsed 4m30s, eta 12m9s"
-			", stalled " + params.stalledOn + ", elapsed 4m30s, remain 12m9s",
+
-		} {
 		if !strings.Contains(shown, want) {
 			t.Errorf("terminal shows %q, want %q", shown, want)
 		}
-		}
 
-		shown = drawStalled(params, 80, 9876543, 1234567, 150*time.Minute)
+		shown = drawStalled(params, 9876543, 1234567, 150*time.Minute)
 		for _, want := range []string{
 			params.label + ": [1234567/" + approx + "9876543] 12%, ",
-			", elapsed 2h30m0s, remain 17h30m0s",
+			", elapsed 2h30m0s, eta 17h30m0s",
 		} {
 			if !strings.Contains(shown, want) {
 				t.Errorf("terminal shows %q, want %q", shown, want)
@@ -691,7 +693,7 @@ func TestTerminalLinesFitWidth(t *testing.T) {
 	for _, width := range []int{120, 100, 80, 72, 60, 50, 40} {
 		for _, total := range []int64{9876543, 20000} {
 			params := progressParams{
-				label: scanPhase("update"), unit: scanUnit, total: total,
+				label: scanPhase("content"), unit: scanUnit, total: total,
 			}
 			out := drawLate(stderr, params, width, total/8)
 
@@ -713,164 +715,13 @@ func TestTerminalLinesFitWidth(t *testing.T) {
 	}
 }
 
-// TestTerminalLinesStartWithSpinner draws a display with no total and
-// one with a bar, two and a half hours into the phase, on a terminal 120
-// columns wide. Each line must start with the library's spinner in the
-// first column, then a space and the phase's label.
-//
-//nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
-func TestTerminalLinesStartWithSpinner(t *testing.T) {
-	stderr := captureStderr(t)
-
-	for _, params := range []progressParams{
-		{label: scanPhase("walk"), unit: scanUnit, total: -1},
-		{label: scanPhase("hash"), unit: scanUnit, total: 9876543},
-	} {
-		shown := lastFrame(drawLate(stderr, params, 120, 1234567))
-		spinner, rest, _ := strings.Cut(shown, " ")
-
-		if utf8.RuneCountInString(spinner) != 1 ||
-			!strings.HasPrefix(rest, params.label+": ") {
-			t.Errorf("terminal shows %q, want the spinner, a space and %q",
-				shown, params.label)
-		}
-
-		if params.total > 0 && !strings.Contains(shown, "% |") {
-			t.Errorf("terminal shows %q, want a bar", shown)
-		}
-	}
-}
-
-// TestTerminalBarStaysAtRightEdge draws the hash phase's display two
-// and a half hours in, at two counts whose lines differ in length, on
-// terminals 110 and 120 columns wide. The bar is partly filled at the
-// first count and full at the second. It must start the same number of
-// columns from the terminal's right edge each time, and end the same
-// number of columns from it: a line that grows or shrinks, as its time
-// remaining does, never moves the bar, and nor do the bar's filled
-// cells.
-//
-//nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
-func TestTerminalBarStaysAtRightEdge(t *testing.T) {
-	stderr := captureStderr(t)
-	params := progressParams{
-		label: scanPhase("hash"), unit: scanUnit, total: 9876543,
-	}
-
-	widths := []int{110, 120}
-	counts := []int64{1234567, 9876543}
-	startFromEdge := make([]int, 0, len(widths)*len(counts))
-	endFromEdge := make([]int, 0, len(widths)*len(counts))
-
-	for _, width := range widths {
-		for _, count := range counts {
-			shown := lastFrame(drawLate(stderr, params, width, count))
-
-			beforeBar, _, found := strings.Cut(shown, "% |")
-			if !found {
-				t.Fatalf("terminal %d wide shows %q, want a bar", width, shown)
-			}
-
-			// The bar ends with the last "|".
-			throughBar := shown[:strings.LastIndex(shown, "|")+1]
-
-			startFromEdge = append(startFromEdge,
-				width-utf8.RuneCountInString(beforeBar))
-			endFromEdge = append(endFromEdge,
-				width-utf8.RuneCountInString(throughBar))
-		}
-	}
-
-	if slices.Min(startFromEdge) != slices.Max(startFromEdge) {
-		t.Errorf("bar starts %v columns from the right edge, want the same "+
-			"every time", startFromEdge)
-	}
-
-	if slices.Min(endFromEdge) != slices.Max(endFromEdge) {
-		t.Errorf("bar ends %v columns from the right edge, want the same "+
-			"every time", endFromEdge)
-	}
-}
-
-// TestTerminalRateGivesWayBeforeBar draws the hash phase's display two
-// and a half hours in. On a terminal 90 columns wide the line fits
-// beside the bar without its rate, but not with it: the bar is drawn
-// and the rate is not. On a terminal 85 columns wide the line does not
-// fit beside the bar even without its rate: neither is drawn, though
-// the rate would fit where the bar was.
-//
-//nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
-func TestTerminalRateGivesWayBeforeBar(t *testing.T) {
-	stderr := captureStderr(t)
-	params := progressParams{
-		label: scanPhase("hash"), unit: scanUnit, total: 9876543,
-	}
-
-	shown := lastFrame(drawLate(stderr, params, 90, 1234567))
-	if !strings.Contains(shown, "% |") || strings.Contains(shown, "files/s") {
-		t.Errorf("terminal 90 wide shows %q, want the bar and no rate", shown)
-	}
-
-	shown = lastFrame(drawLate(stderr, params, 85, 1234567))
-	if strings.Contains(shown, "% |") || strings.Contains(shown, "files/s") {
-		t.Errorf("terminal 85 wide shows %q, want neither bar nor rate", shown)
-	}
-}
-
-// TestTerminalSpinnerTurnsBesideBar draws the hash phase's display, its
-// bar partly filled, several times on a terminal 120 columns wide. The
-// spinner must turn at every draw. A filled cell of the bar takes one
-// column but three bytes. A line measured in bytes would seem too wide
-// for the terminal, and every draw would start the display over, its
-// spinner included.
-//
-//nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
-func TestTerminalSpinnerTurnsBesideBar(t *testing.T) {
-	stderr := captureStderr(t)
-	params := progressParams{
-		label: scanPhase("hash"), unit: scanUnit, total: 9876543,
-	}
-
-	// No timed draws: the test draws.
-	p := startProgress(params, func() int { return 120 }, time.Hour,
-		time.Now)
-
-	p.mu.Lock()
-	p.count = 1234567
-	p.mu.Unlock()
-
-	const draws = 4
-
-	spinners := make([]string, 0, draws)
-
-	for range draws {
-		before := len(stderr())
-
-		p.draw()
-
-		spinner, _, _ := strings.Cut(lastFrame(stderr()[before:]), " ")
-		spinners = append(spinners, spinner)
-	}
-
-	p.finish()
-
-	for i := 1; i < len(spinners); i++ {
-		if spinners[i] == spinners[i-1] {
-			t.Errorf("spinners at successive draws are %q, want each to "+
-				"differ from the one before", spinners)
-
-			break
-		}
-	}
-}
-
 // TestTerminalNarrowedRedrawsInPlace draws a large scan's display, two
 // and a half hours into the phase, on a terminal 120 columns wide, then
 // makes the terminal narrower than that line and draws it again a few
 // times. Nothing written after that, the erase before each redraw
 // included, may reach the narrower terminal's last column: it would
-// wrap, and every redraw would leave a line behind. At the narrower
+// wrap, and every redraw would leave a line behind. At these widths the
-// width the hash phase's bar gives way.
+// spinner stays a spinner and the bar gives way to the spinner.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
 func TestTerminalNarrowedRedrawsInPlace(t *testing.T) {
@@ -888,7 +739,7 @@ func TestTerminalNarrowedRedrawsInPlace(t *testing.T) {
 			progressParams{
 				label: scanPhase("hash"), unit: scanUnit, total: 9876543,
 			},
-			80,
+			90,
 		},
 	}
 
@@ -928,10 +779,10 @@ func TestTerminalNarrowedRedrawsInPlace(t *testing.T) {
 	}
 }
 
-// TestProgressWarningsOnOwnLines drives the terminal displays, with a
+// TestProgressWarningsOnOwnLines drives the terminal displays, spinner
-// bar and without, through a run of warnings with no items between
+// and bar, through a run of warnings with no items between them, as when
-// them, as when the walk meets many unreadable paths, for many redraws:
+// the walk meets many unreadable paths, for many redraws: every warning
-// every warning must land on a line of its own, never inside a redraw.
+// must land on a line of its own, never inside a redraw.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
 func TestProgressWarningsOnOwnLines(t *testing.T) {
@@ -977,12 +828,11 @@ func TestProgressWarningsOnOwnLines(t *testing.T) {
 	}
 }
 
-// TestProgressWriterRowsOnOwnLines drives the terminal displays, with a
+// TestProgressWriterRowsOnOwnLines drives the terminal displays,
-// bar and without, through a run of rows written through the display's
+// spinner and bar, through a run of rows written through the display's
 // writer to the same terminal, as stdout is in a plain "sfdupes trees",
 // each written out at once, for many redraws: the terminal must show
-// every row whole, in order, on a line of its own. The terminal is 120
+// every row whole, in order, on a line of its own.
-// columns wide, so the write phase's display has room for its bar.
 //
 //nolint:paralleltest // captureStderr replaces the process-wide os.Stderr
 func TestProgressWriterRowsOnOwnLines(t *testing.T) {
@@ -999,8 +849,7 @@ func TestProgressWriterRowsOnOwnLines(t *testing.T) {
 	var want []string
 
 	for _, params := range cases {
-		p := startProgress(params, func() int { return 120 }, testRedraw,
+		p := startProgress(params, terminalWidth, testRedraw, time.Now)
-			time.Now)
 		out := p.writer(os.Stderr)
 
 		// No pause between rows: one is garbled only if a redraw lands

report.go

@@ -2,7 +2,6 @@ package main
 
 import (
 	"context"
-	"crypto/sha256"
 	"database/sql"
 	"fmt"
 	"io"
@@ -25,9 +24,9 @@ const minGroupSize = 2
 type scanRec struct {
 	size    int64
 	mtime   time.Time
-	head    [sha256.Size]byte
+	head    string
-	tail    [sha256.Size]byte
+	tail    string
-	content [sha256.Size]byte
+	content string
 	path    string
 }
 
@@ -170,6 +169,12 @@ type reportWriter struct {
 func (w *reportWriter) add(r scanRec) error {
 	w.records++
 
+	// A record without a content hash has unknown content (README
+	// "Database"), so it is never a duplicate.
+	if r.content == "" {
+		return nil
+	}
+
 	if r.size == w.sig.size && r.head == w.sig.head &&
 		r.tail == w.sig.tail && r.content == w.sig.content {
 		w.paths = append(w.paths, r.path)

report_test.go

@@ -2,7 +2,6 @@ package main
 
 import (
 	"bytes"
-	"crypto/sha256"
 	"database/sql"
 	"errors"
 	"fmt"
@@ -19,27 +18,14 @@ import (
 // awkwardDir is a directory name holding every byte the reports escape.
 const awkwardDir = "/d/\tone\ntwo\rthree\\four"
 
-// testHash returns a stand-in hash for a test record: the bytes of s,
-// then zeros. Stand-ins sort as their strings do, so a test can choose
-// the order in which report reads its records.
-func testHash(s string) [sha256.Size]byte {
-	var h [sha256.Size]byte
-
-	copy(h[:], s)
-
-	return h
-}
-
 // awkwardPairRecs is a duplicate pair in sibling directories /d/A and
 // awkwardDir. A raw tab sorts before "A" but its escaped form `\t`
 // sorts after it, so awkwardDir coming first shows that sorting uses
 // the raw path.
 func awkwardPairRecs() []scanRec {
-	h, tl, c := testHash("h"), testHash("t"), testHash("c")
-
 	return []scanRec{
-		{size: 5, head: h, tail: tl, content: c, path: "/d/A/f"},
+		{size: 5, head: "h", tail: "t", content: "c", path: "/d/A/f"},
-		{size: 5, head: h, tail: tl, content: c, path: awkwardDir + "/f"},
+		{size: 5, head: "h", tail: "t", content: "c", path: awkwardDir + "/f"},
 	}
 }
 
@@ -157,8 +143,8 @@ func TestReportStdoutFailsWhileReading(t *testing.T) {
 	recs := make([]scanRec, 2*ioBufSize/len(dir))
 	for i := range recs {
 		recs[i] = scanRec{
-			size: int64(i/2) + 1, head: testHash("h"), tail: testHash("t"),
+			size: int64(i/2) + 1, head: "h", tail: "t", content: "c",
-			content: testHash("c"), path: fmt.Sprintf("%s/%d", dir, i),
+			path: fmt.Sprintf("%s/%d", dir, i),
 		}
 	}
 
@@ -187,7 +173,7 @@ func TestReportStdoutFailsWhileReading(t *testing.T) {
 // reportOpenLine is the plain line of report's open phase when it opens
 // the database at once; it prints it as the phase starts and as it ends.
 const reportOpenLine = "[1/2] open: opening the database, elapsed 0s, " +
-	"remain unknown"
+	"eta unknown"
 
 // smokeReport is report's stdout for the smoke-test records.
 const smokeReport = "first\tdupe\tsize\n" +
@@ -225,7 +211,7 @@ func TestRunReportProgressNamesEachPhase(t *testing.T) {
 		"report: database " + path + " (from " + databaseEnv + ")",
 		reportOpenLine,
 		reportOpenLine,
-		"[2/2] read: counting the records, elapsed 0s, remain unknown",
+		"[2/2] read: counting the records, elapsed 0s, eta unknown",
 		"[2/2] read: [6/6] 100%, ",
 		"report: 6 records read, 2 duplicate groups, 4 dupe files, " +
 			"6.1 KiB reclaimable",
@@ -291,14 +277,13 @@ func TestReadReportOnTerminal(t *testing.T) {
 func TestRunReportsIgnoreInsertionOrder(t *testing.T) {
 	// README §Constraints: identical database contents give identical
 	// output, whatever order the records were inserted in.
-	a, b := testHash("a"), testHash("b")
-
 	recs := append(smokeTreeRecs(), awkwardPairRecs()...)
 	recs = append(recs,
-		scanRec{size: 50, head: b, tail: b, content: b, path: "/y/2"},
+		scanRec{size: 50, head: "b", tail: "b", content: "b", path: "/y/2"},
-		scanRec{size: 50, head: b, tail: b, content: b, path: "/y/1"},
+		scanRec{size: 50, head: "b", tail: "b", content: "b", path: "/y/1"},
-		scanRec{size: 50, head: a, tail: a, content: a, path: "/x/2"},
+		scanRec{size: 50, head: "a", tail: "a", content: "a", path: "/x/2"},
-		scanRec{size: 50, head: a, tail: a, content: a, path: "/x/1"},
+		scanRec{size: 50, head: "a", tail: "a", content: "a", path: "/x/1"},
+		scanRec{size: 50, path: "/x/unhashed"},
 	)
 
 	reversed := slices.Clone(recs)
@@ -401,19 +386,16 @@ func TestWarnfEscapes(t *testing.T) {
 func TestDupeGroups(t *testing.T) {
 	t.Parallel()
 
-	h, tl, c := testHash("h"), testHash("t"), testHash("c")
-	bigH, bigT, bigC := testHash("H"), testHash("T"), testHash("C")
-	u := testHash("u")
 	recs := []scanRec{
-		{size: 100, head: h, tail: tl, content: c, path: "/z/b"},
+		{size: 100, head: "h", tail: "t", content: "c", path: "/z/b"},
-		{size: 100, head: h, tail: tl, content: c, path: "/z/a"},
+		{size: 100, head: "h", tail: "t", content: "c", path: "/z/a"},
-		{size: 100, head: h, tail: tl, content: c, path: "/z/c"},
+		{size: 100, head: "h", tail: "t", content: "c", path: "/z/c"},
-		{size: 4000, head: bigH, tail: bigT, content: bigC, path: "/big/2"},
+		{size: 4000, head: "H", tail: "T", content: "C", path: "/big/2"},
-		{size: 4000, head: bigH, tail: bigT, content: bigC, path: "/big/1"},
+		{size: 4000, head: "H", tail: "T", content: "C", path: "/big/1"},
 		// Same size as the /z group but a different head hash.
-		{size: 100, head: testHash("other"), tail: tl, content: c, path: "/z/d"},
+		{size: 100, head: "other", tail: "t", content: "c", path: "/z/d"},
 		// A singleton signature must not form a group.
-		{size: 7, head: u, tail: u, content: u, path: "/lonely"},
+		{size: 7, head: "u", tail: "u", content: "u", path: "/lonely"},
 	}
 
 	groups := dupeGroupsOf(t, recs)
@@ -437,13 +419,16 @@ func TestDupeGroups(t *testing.T) {
 func TestDupeGroupsContentSeparates(t *testing.T) {
 	t.Parallel()
 
-	// Same size, head, and tail: /b differs from the others only in its
+	// Same size, head, and tail, but different content hashes: the final
-	// content hash, so it joins no group, while /a and /c group.
+	// rung keeps them apart, so no group forms. Matching content groups.
-	h, tl := testHash("h"), testHash("t")
+	// Records without a content hash never group, not even with each
+	// other.
 	recs := []scanRec{
-		{size: 100, head: h, tail: tl, content: testHash("c1"), path: "/a"},
+		{size: 100, head: "h", tail: "t", content: "c1", path: "/a"},
-		{size: 100, head: h, tail: tl, content: testHash("c2"), path: "/b"},
+		{size: 100, head: "h", tail: "t", content: "c2", path: "/b"},
-		{size: 100, head: h, tail: tl, content: testHash("c1"), path: "/c"},
+		{size: 100, head: "h", tail: "t", content: "c1", path: "/c"},
+		{size: 100, head: "h", tail: "t", path: "/d"},
+		{size: 100, head: "h", tail: "t", path: "/e"},
 	}
 
 	groups := dupeGroupsOf(t, recs)
@@ -462,15 +447,14 @@ func TestDupeGroupsMtimeExcluded(t *testing.T) {
 
 	// mtime is informational only; records differing only in mtime
 	// still group together.
-	h, tl, c := testHash("h"), testHash("t"), testHash("c")
 	recs := []scanRec{
 		{
-			size: 9, mtime: time.Unix(100, 0), head: h, tail: tl,
+			size: 9, mtime: time.Unix(100, 0), head: "h", tail: "t",
-			content: c, path: "/m/1",
+			content: "c", path: "/m/1",
 		},
 		{
-			size: 9, mtime: time.Unix(200, 0), head: h, tail: tl,
+			size: 9, mtime: time.Unix(200, 0), head: "h", tail: "t",
-			content: c, path: "/m/2",
+			content: "c", path: "/m/2",
 		},
 	}
 
@@ -485,12 +469,11 @@ func TestDupeGroupsTieBreak(t *testing.T) {
 
 	// The hashes sort opposite to the first paths, so ordering the
 	// groups by hash instead of by first path fails this test.
-	a, b := testHash("a"), testHash("b")
 	recs := []scanRec{
-		{size: 50, head: a, tail: a, content: a, path: "/beta/2"},
+		{size: 50, head: "a", tail: "a", content: "a", path: "/beta/2"},
-		{size: 50, head: a, tail: a, content: a, path: "/beta/1"},
+		{size: 50, head: "a", tail: "a", content: "a", path: "/beta/1"},
-		{size: 50, head: b, tail: b, content: b, path: "/alpha/2"},
+		{size: 50, head: "b", tail: "b", content: "b", path: "/alpha/2"},
-		{size: 50, head: b, tail: b, content: b, path: "/alpha/1"},
+		{size: 50, head: "b", tail: "b", content: "b", path: "/alpha/1"},
 	}
 
 	groups := dupeGroupsOf(t, recs)
@@ -512,16 +495,15 @@ func TestDupeGroupsOrderedByFirstPath(t *testing.T) {
 	// /z, /k, which is neither the order of their first paths nor its
 	// reverse, and must write them ordered by first path. The group of
 	// size 7 comes after them all, though its first path sorts first.
-	a, b, c, d := testHash("a"), testHash("b"), testHash("c"), testHash("d")
 	recs := []scanRec{
-		{size: 50, head: a, tail: a, content: a, path: "/k/2"},
+		{size: 50, head: "a", tail: "a", content: "a", path: "/k/2"},
-		{size: 50, head: a, tail: a, content: a, path: "/k/1"},
+		{size: 50, head: "a", tail: "a", content: "a", path: "/k/1"},
-		{size: 50, head: b, tail: b, content: b, path: "/z/1"},
+		{size: 50, head: "b", tail: "b", content: "b", path: "/z/1"},
-		{size: 50, head: b, tail: b, content: b, path: "/z/2"},
+		{size: 50, head: "b", tail: "b", content: "b", path: "/z/2"},
-		{size: 50, head: c, tail: c, content: c, path: "/a/2"},
+		{size: 50, head: "c", tail: "c", content: "c", path: "/a/2"},
-		{size: 50, head: c, tail: c, content: c, path: "/a/1"},
+		{size: 50, head: "c", tail: "c", content: "c", path: "/a/1"},
-		{size: 7, head: d, tail: d, content: d, path: "/0/2"},
+		{size: 7, head: "d", tail: "d", content: "d", path: "/0/2"},
-		{size: 7, head: d, tail: d, content: d, path: "/0/1"},
+		{size: 7, head: "d", tail: "d", content: "d", path: "/0/1"},
 	}
 
 	groups := dupeGroupsOf(t, recs)
@@ -540,12 +522,11 @@ func TestDupeGroupsOrderedByFirstPath(t *testing.T) {
 func TestDupeGroupsDeterministic(t *testing.T) {
 	t.Parallel()
 
-	a, b := testHash("a"), testHash("b")
 	recs := []scanRec{
-		{size: 1, head: a, tail: a, content: a, path: "/p/1"},
+		{size: 1, head: "a", tail: "a", content: "a", path: "/p/1"},
-		{size: 1, head: a, tail: a, content: a, path: "/p/2"},
+		{size: 1, head: "a", tail: "a", content: "a", path: "/p/2"},
-		{size: 2, head: b, tail: b, content: b, path: "/q/1"},
+		{size: 2, head: "b", tail: "b", content: "b", path: "/q/1"},
-		{size: 2, head: b, tail: b, content: b, path: "/q/2"},
+		{size: 2, head: "b", tail: "b", content: "b", path: "/q/2"},
 	}
 
 	forward := dupeGroupsOf(t, recs)

trees.go

@@ -286,9 +286,18 @@ func (b *treeBuilder) finish() (*treeNode, []*treeNode) {
 // name and its signature (size, head, tail, content); mtime is
 // excluded.
 func fileEntry(name string, r scanRec) string {
+	content := r.content
+
+	// A record without a content hash has unknown content (README
+	// "Database"): give it a signature no other file can share, so
+	// trees containing it never compare equal. Real hashes are hex, so
+	// the NUL-prefixed form cannot collide.
+	if content == "" {
+		content = "unhashed\x00" + r.path
+	}
+
 	return "f\x00" + name + "\x00" + strconv.FormatInt(r.size, 10) +
-		"\x00" + string(r.head[:]) + "\x00" + string(r.tail[:]) +
+		"\x00" + r.head + "\x00" + r.tail + "\x00" + content
-		"\x00" + string(r.content[:])
 }
 
 // collectTreeGroups groups directories by digest, counting each one on
@@ -338,8 +347,7 @@ func collectTreeGroups(allDirs []*treeNode, super *treeNode,
 // digest is the SHA-256 of its child entries — files serialized with
 // name and signature, subdirectories with name and recursive digest —
 // sorted byte-lexicographically. Filenames cannot contain NUL or "/",
-// and every hash and digest is 32 bytes, so NUL delimiters are
+// so NUL delimiters are unambiguous.
-// unambiguous.
 func (n *treeNode) computeDigest() {
 	slices.Sort(n.entries)
 

trees_test.go

@@ -10,17 +10,24 @@ import (
 	"time"
 )
 
+// Signature hashes shared by the smoke-test records.
+const (
+	f1Head    = "f1h"
+	f1Tail    = "f1t"
+	f1Content = "f1c"
+	f2Head    = "f2h"
+	f2Tail    = "f2t"
+	f2Content = "f2c"
+)
+
 // treesOpenLine is the plain line of trees' open phase when it opens
 // the database at once; it prints it as the phase starts and as it ends.
 const treesOpenLine = "[1/4] open: opening the database, elapsed 0s, " +
-	"remain unknown"
+	"eta unknown"
 
 // smokeTreeRecs mirrors the README smoke-test tree layout: /d/t1 and
 // /d/t2 are identical, /d/t3 differs from them only by one filename.
 func smokeTreeRecs() []scanRec {
-	f1Head, f1Tail, f1Content := testHash("f1h"), testHash("f1t"), testHash("f1c")
-	f2Head, f2Tail, f2Content := testHash("f2h"), testHash("f2t"), testHash("f2c")
-
 	return []scanRec{
 		{size: 3000, head: f1Head, tail: f1Tail, content: f1Content, path: "/d/t1/f1"},
 		{size: 100, head: f2Head, tail: f2Tail, content: f2Content, path: "/d/t1/sub/f2"},
@@ -146,14 +153,13 @@ func TestTreeNamesSortingBeforeSlash(t *testing.T) {
 
 	for _, top := range []string{"/a", "/c"} {
 		for _, p := range []string{"/b", "/b-x/f", "/b.txt", "/b/f"} {
-			content := testHash("c")
+			content := "c"
 			if p == "/b-x/f" {
-				content = testHash("other")
+				content = "other"
 			}
 
 			recs = append(recs, scanRec{
-				size: 1, head: testHash("h"), tail: testHash("t"),
+				size: 1, head: "h", tail: "t", content: content, path: top + p,
-				content: content, path: top + p,
 			})
 		}
 	}
@@ -236,11 +242,11 @@ func TestRunTreesProgressNamesEachPhase(t *testing.T) {
 		"trees: database " + path + " (from " + databaseEnv + ")",
 		treesOpenLine,
 		treesOpenLine,
-		"[2/4] read: counting the records, elapsed 0s, remain unknown",
+		"[2/4] read: counting the records, elapsed 0s, eta unknown",
 		"[2/4] read: [6/6] 100%, ",
-		"[3/4] group: [0/8] 0%, 0 directories/s, elapsed 0s, remain unknown",
+		"[3/4] group: [0/8] 0%, 0 directories/s, elapsed 0s, eta unknown",
 		"[3/4] group: [8/8] 100%, ",
-		"[4/4] write: [0/1] 0%, 0 groups/s, elapsed 0s, remain unknown",
+		"[4/4] write: [0/1] 0%, 0 groups/s, elapsed 0s, eta unknown",
 		"[4/4] write: [1/1] 100%, ",
 		"trees: 6 records read, 1 duplicate tree groups, 1 dupe trees, ",
 	}
@@ -326,10 +332,11 @@ func TestTreeDigests(t *testing.T) {
 func TestTreeDigestContentSensitivity(t *testing.T) {
 	t.Parallel()
 
-	same, c := testHash("same"), testHash("c")
+	const sharedTail = "same"
+
 	recs := []scanRec{
-		{size: 10, head: same, tail: same, content: c, path: "/r/a/f"},
+		{size: 10, head: sharedTail, tail: sharedTail, content: "c", path: "/r/a/f"},
-		{size: 10, head: testHash("DIFF"), tail: same, content: c, path: "/r/b/f"},
+		{size: 10, head: "DIFF", tail: sharedTail, content: "c", path: "/r/b/f"},
 	}
 
 	_, dirs := treeOf(t, recs)
@@ -390,10 +397,9 @@ func TestCollectTreeGroupsSiblings(t *testing.T) {
 
 	// Identical sibling dirs share a parent, so their group cannot be
 	// implied by a parent group and must be reported.
-	h, tl, c := testHash("h"), testHash("t"), testHash("c")
 	recs := []scanRec{
-		{size: 10, head: h, tail: tl, content: c, path: "/p/x1/f"},
+		{size: 10, head: "h", tail: "t", content: "c", path: "/p/x1/f"},
-		{size: 10, head: h, tail: tl, content: c, path: "/p/x2/f"},
+		{size: 10, head: "h", tail: "t", content: "c", path: "/p/x2/f"},
 	}
 
 	super, dirs := treeOf(t, recs)
@@ -412,11 +418,10 @@ func TestCollectTreeGroupsDifferingParents(t *testing.T) {
 	// /p/a and /q/b contain identical x subtrees, but /p/a has an
 	// extra file, so the parents' digests differ and the x group must
 	// be reported.
-	h, tl, c, e := testHash("h"), testHash("t"), testHash("c"), testHash("e")
 	recs := []scanRec{
-		{size: 10, head: h, tail: tl, content: c, path: "/p/a/x/f"},
+		{size: 10, head: "h", tail: "t", content: "c", path: "/p/a/x/f"},
-		{size: 99, head: e, tail: e, content: e, path: "/p/a/extra"},
+		{size: 99, head: "e", tail: "e", content: "e", path: "/p/a/extra"},
-		{size: 10, head: h, tail: tl, content: c, path: "/q/b/x/f"},
+		{size: 10, head: "h", tail: "t", content: "c", path: "/q/b/x/f"},
 	}
 
 	super, dirs := treeOf(t, recs)