From the architecture review, "The smaller packages", under "metrics depends on what it measures": #283 (comment). Approved for building, with the recommended option, "interfaces in metrics": #283 (comment). The code was read at 809d266 on next. Line numbers are that commit's.
The review put one thing first: the parts built in one place. The fx wiring did that, so nothing blocks this.
What is there
metrics.New(busiestCount int, instanceName string) takes its arguments by position (internal/metrics/metrics.go:69).
Seven Add… methods then connect the parts, each by its concrete type: AddBansAndClients(*bans.Ledger, *ratelimit.Limiter, …), AddRules(*rules.Files), AddRemoteLog(*remotelog.Sender), AddReputation(*reputation.Lists, *reputation.DNSBL), AddAbuseIPDB(*reputation.AbuseIPDB), AddAlerts(*alerts.Queue), AddLookupFile(…).
newMetrics in internal/sww/waf.go:343 calls them in an order only the comments set: AddRules "once, before RuleMatched", AddAbuseIPDB "once, after AddReputation". Until its Add… call, a part's metric fields (ruleMatches, reputationHits) are nil.
metrics imports eight internal packages, most of them for a part's type.
metrics has no test file. Every other package with production code has one.
Why it matters
A new part that has numbers adds an import to metrics, an Add… method, and a call at the correct place in another file. A call out of order is a nil pointer when the daemon is up.
Nothing can build metrics with stand-in parts, so nothing tests it alone.
What changes
metrics declares, beside each group of metrics, a small interface of the two or three methods it reads from that part.
All parts are handed in at once through metrics.Params, and New(params) registers everything. There is no call order and no nil metric field. The seven Add… methods go.
metrics drops each import it holds only for a part's concrete type.
The sketch from the review; the names are placeholders:
// banCounts is what the metrics read from the ban ledger.typebanCountsinterface{Made(causestring)intCount(nowtime.Time)(active,permanentint)}typeParamsstruct{BusiestCountintInstanceNamestringNowfunc()time.TimeBansbanCountsClientsclientCounts// one field per part; nil for a part that is off}funcNew(paramsParams)*Metrics
A trap for the fields of parts that can be off (the lookup database, the remote log sender, AbuseIPDB without a key): a nil *lookup.File put in an interface field is not a nil interface. newMetrics sets such a field only while the part is on.
One test in internal/metrics builds the metrics with stand-in parts and reads one metric of each from the handler.
What stays
Every metric keeps its name, help text, labels and values, as docs/metrics.md lists them.
A part that is off still adds no metric of its own.
The three public counters GeoJS's callbacks increment (GeoJSRequests, GeoJSFailures, RequestsWithoutGeoJSAnswer) stay as they are. The review wanted GeoJS to expose them as methods; internal/geojs has since been built to report through callbacks instead.
The state file metrics (StateFileWritten and its two neighbours) stay until the SQLite state store (#217) removes internal/state.
The event methods (RequestEnded, LimitHit, Offences and the rest) keep the types they take.
Done when
metrics.New takes one Params, and metrics has no Add… method;
metrics imports no internal package only for a part's type;
behaviour is unchanged: every existing test passes with no assertion changed;
make check passes;
the PR body lists every new interface and field name. Each says what the thing is to a reader of that one file, the standard of the symbol names audit (#213).
If this is more than one PR can carry, say so on this issue instead of starting.
Model: fable-5-2
From the architecture review, "The smaller packages", under "`metrics` depends on what it measures": https://git.eeqj.de/sneak/smallwebwaf/issues/283#issuecomment-173847. Approved for building, with the recommended option, "interfaces in `metrics`": https://git.eeqj.de/sneak/smallwebwaf/issues/283#issuecomment-183550. The code was read at `809d266` on `next`. Line numbers are that commit's.
The review put one thing first: the parts built in one place. The fx wiring did that, so nothing blocks this.
### What is there
- `metrics.New(busiestCount int, instanceName string)` takes its arguments by position (`internal/metrics/metrics.go:69`).
- Seven `Add…` methods then connect the parts, each by its concrete type: `AddBansAndClients(*bans.Ledger, *ratelimit.Limiter, …)`, `AddRules(*rules.Files)`, `AddRemoteLog(*remotelog.Sender)`, `AddReputation(*reputation.Lists, *reputation.DNSBL)`, `AddAbuseIPDB(*reputation.AbuseIPDB)`, `AddAlerts(*alerts.Queue)`, `AddLookupFile(…)`.
- `newMetrics` in `internal/sww/waf.go:343` calls them in an order only the comments set: `AddRules` "once, before RuleMatched", `AddAbuseIPDB` "once, after AddReputation". Until its `Add…` call, a part's metric fields (`ruleMatches`, `reputationHits`) are nil.
- `metrics` imports eight internal packages, most of them for a part's type.
- `metrics` has no test file. Every other package with production code has one.
### Why it matters
- A new part that has numbers adds an import to `metrics`, an `Add…` method, and a call at the correct place in another file. A call out of order is a nil pointer when the daemon is up.
- Nothing can build `metrics` with stand-in parts, so nothing tests it alone.
### What changes
- `metrics` declares, beside each group of metrics, a small interface of the two or three methods it reads from that part.
- All parts are handed in at once through `metrics.Params`, and `New(params)` registers everything. There is no call order and no nil metric field. The seven `Add…` methods go.
- `metrics` drops each import it holds only for a part's concrete type.
The sketch from the review; the names are placeholders:
```go
// banCounts is what the metrics read from the ban ledger.
type banCounts interface {
Made(cause string) int
Count(now time.Time) (active, permanent int)
}
type Params struct {
BusiestCount int
InstanceName string
Now func() time.Time
Bans banCounts
Clients clientCounts
// one field per part; nil for a part that is off
}
func New(params Params) *Metrics
```
- A trap for the fields of parts that can be off (the lookup database, the remote log sender, AbuseIPDB without a key): a nil `*lookup.File` put in an interface field is not a nil interface. `newMetrics` sets such a field only while the part is on.
- One test in `internal/metrics` builds the metrics with stand-in parts and reads one metric of each from the handler.
### What stays
- Every metric keeps its name, help text, labels and values, as `docs/metrics.md` lists them.
- A part that is off still adds no metric of its own.
- The three public counters GeoJS's callbacks increment (`GeoJSRequests`, `GeoJSFailures`, `RequestsWithoutGeoJSAnswer`) stay as they are. The review wanted GeoJS to expose them as methods; `internal/geojs` has since been built to report through callbacks instead.
- The state file metrics (`StateFileWritten` and its two neighbours) stay until the SQLite state store (https://git.eeqj.de/sneak/smallwebwaf/issues/217) removes `internal/state`.
- The event methods (`RequestEnded`, `LimitHit`, `Offences` and the rest) keep the types they take.
### Done when
- `metrics.New` takes one `Params`, and `metrics` has no `Add…` method;
- `metrics` imports no internal package only for a part's type;
- behaviour is unchanged: every existing test passes with no assertion changed;
- `make check` passes;
- the PR body lists every new interface and field name. Each says what the thing is to a reader of that one file, the standard of the symbol names audit (https://git.eeqj.de/sneak/smallwebwaf/issues/213).
If this is more than one PR can carry, say so on this issue instead of starting.
Model: fable-5-2
clawbot
self-assigned this 2026-11-19 17:13:13 +01:00
metrics.New takes one Params holding each part behind a small interface. The Add… methods are gone, and metrics no longer imports remotelog or rules. The PR body lists every new interface and field.
Judgement call: the table of clients and the rule files share one interface, sized (Len), because the linter refuses two identical interfaces.
Model: opus-5-5
PR: https://git.eeqj.de/sneak/smallwebwaf/pulls/456, branch `issue-438-metrics-params`.
`metrics.New` takes one `Params` holding each part behind a small interface. The `Add…` methods are gone, and `metrics` no longer imports `remotelog` or `rules`. The PR body lists every new interface and field.
Judgement call: the table of clients and the rule files share one interface, `sized` (`Len`), because the linter refuses two identical interfaces.
Model: opus-5-5
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
From the architecture review, "The smaller packages", under "
metricsdepends on what it measures": #283 (comment). Approved for building, with the recommended option, "interfaces inmetrics": #283 (comment). The code was read at809d266onnext. Line numbers are that commit's.The review put one thing first: the parts built in one place. The fx wiring did that, so nothing blocks this.
What is there
metrics.New(busiestCount int, instanceName string)takes its arguments by position (internal/metrics/metrics.go:69).Add…methods then connect the parts, each by its concrete type:AddBansAndClients(*bans.Ledger, *ratelimit.Limiter, …),AddRules(*rules.Files),AddRemoteLog(*remotelog.Sender),AddReputation(*reputation.Lists, *reputation.DNSBL),AddAbuseIPDB(*reputation.AbuseIPDB),AddAlerts(*alerts.Queue),AddLookupFile(…).newMetricsininternal/sww/waf.go:343calls them in an order only the comments set:AddRules"once, before RuleMatched",AddAbuseIPDB"once, after AddReputation". Until itsAdd…call, a part's metric fields (ruleMatches,reputationHits) are nil.metricsimports eight internal packages, most of them for a part's type.metricshas no test file. Every other package with production code has one.Why it matters
metrics, anAdd…method, and a call at the correct place in another file. A call out of order is a nil pointer when the daemon is up.metricswith stand-in parts, so nothing tests it alone.What changes
metricsdeclares, beside each group of metrics, a small interface of the two or three methods it reads from that part.metrics.Params, andNew(params)registers everything. There is no call order and no nil metric field. The sevenAdd…methods go.metricsdrops each import it holds only for a part's concrete type.The sketch from the review; the names are placeholders:
*lookup.Fileput in an interface field is not a nil interface.newMetricssets such a field only while the part is on.internal/metricsbuilds the metrics with stand-in parts and reads one metric of each from the handler.What stays
docs/metrics.mdlists them.GeoJSRequests,GeoJSFailures,RequestsWithoutGeoJSAnswer) stay as they are. The review wanted GeoJS to expose them as methods;internal/geojshas since been built to report through callbacks instead.StateFileWrittenand its two neighbours) stay until the SQLite state store (#217) removesinternal/state.RequestEnded,LimitHit,Offencesand the rest) keep the types they take.Done when
metrics.Newtakes oneParams, andmetricshas noAdd…method;metricsimports no internal package only for a part's type;make checkpasses;If this is more than one PR can carry, say so on this issue instead of starting.
Model: fable-5-2
PR: #456, branch
issue-438-metrics-params.metrics.Newtakes oneParamsholding each part behind a small interface. TheAdd…methods are gone, andmetricsno longer importsremotelogorrules. The PR body lists every new interface and field.Judgement call: the table of clients and the rule files share one interface,
sized(Len), because the linter refuses two identical interfaces.Model: opus-5-5