datalog-theme 0.10.1 → 0.11.0

data/assets/js/dist/subscriptions.js

@@ -1 +1 @@
-import{a as O,b as R,c as q,d as T,e as I}from"./chunks/chunk-WIRUK3OZ.js";import{a as x,c as C}from"./chunks/chunk-VZ5WKQVA.js";import"./chunks/chunk-PATLC23F.js";var h="subscriptions",M=/^[^\s@]+@[^\s@]+\.[^\s@]+$/;function N(e,a){let s=e.querySelector(a);if(!s)return{};try{return JSON.parse(s.textContent)||{}}catch(n){return{}}}function $(e,a={}){let s={email:e.email||""};return a.hasTopics&&(s.topics=[].concat(e.topics||[]).filter(Boolean)),a.sourceUrl&&(s.source_url=a.sourceUrl),a.locale&&(s.locale=a.locale),s}function j(e,a,s={}){let n={};return M.test(e.email)||(n.email=s.email_invalid||"That does not look like an email address."),Array.isArray(e.topics)&&e.topics.length===0&&(n.topics=s.topics_required||"Pick at least one."),a.website&&(n.website="spam"),Object.keys(n).length>0?n:null}function P(e,a={}){let s=a.client||C(),n=e.querySelector("form"),c=N(e,"[data-subscribe-labels]"),g=c.errors||{},S=e.ownerDocument,o=S.querySelector('link[rel="canonical"]'),k={sourceUrl:e.dataset.subscribeSource||o&&o.href||"",locale:e.dataset.subscribeLocale||S.documentElement.lang||"",hasTopics:n.querySelector('[name="topics"]')!==null},v=e.dataset.subscribeDoubleOptIn!=="false",E=O(n),w=d=>{n.dataset.result=d,q(n,"success",c[d]||"")};if(!s.enabled)return q(n,"disabled",g.disabled||""),{form:n,available:!1,submit:async()=>null};let m=R(s,h,n,g);n.addEventListener("focusin",()=>m().catch(()=>{}),{once:!0});let p={form:n,available:!0,async submit(){let d=T(n),f=$(d,k),l=j(f,d,c);if(delete n.dataset.result,l)return l.website?(w("pending"),null):(q(n,"invalid",g.invalid||"",l),null);q(n,"pending",c.sending||"");try{await m()}catch(i){return null}try{let i=await s.post(s.pathFor(h),f,{idempotencyKey:E.key(f)}),u=i&&i.data&&typeof i.data.status=="string"?i.data.status:"";return E.clear(),n.reset(),w(u==="confirmed"||u==="subscribed"||u===""&&!v?"confirmed":"pending"),i}catch(i){return i&&i.kind==="conflict"?w("duplicate"):I(n,i,g),null}}};return n.addEventListener("submit",d=>{d.preventDefault(),p.submit()}),p}function U(e=document){return Array.from(e.querySelectorAll("[data-subscribe]")).map(a=>P(a))}var J=["confirm","unsubscribe","manage"],K=/^[A-Za-z0-9._~-]{8,512}$/;function z(e,a){let s=e.querySelector(a);if(!s)return{};try{return JSON.parse(s.textContent)||{}}catch(n){return{}}}function B(e){let a=new URLSearchParams(e||""),s=J.find(c=>a.has(c));if(!s)return null;let n=(a.get(s)||"").trim();return K.test(n)?{action:s,token:n}:null}function V(e,a={}){let s=e.ownerDocument,n=s.defaultView,c=a.client||C(),g=a.location||n&&n.location||{search:"",pathname:"/"},S=a.history===void 0?n&&n.history:a.history,o=z(e,"[data-manage-labels]"),k=o.errors||{},v=e.querySelector("[data-manage-status]"),E=e.querySelector("[data-manage-idle]"),w=e.querySelector("[data-manage-unsubscribe-panel]"),m=e.querySelector("[data-manage-preferences]"),p=B(g.search),d=p?`${c.pathFor(h)}/${encodeURIComponent(p.token)}`:"",f=null,l=s.createElement("button");l.type="button",l.className="post-tool",l.dataset.manageRetry="",l.textContent=o.retry||"Try again",l.hidden=!0,l.addEventListener("click",()=>f==null?void 0:f()),e.appendChild(l);let i=(r,t="",A=!1)=>{r!=="error"&&(f=null),e.dataset.state=r,l.hidden=r!=="error"||!f,v&&(v.textContent=t,v.hidden=t==="",v.setAttribute("role",A?"alert":"status"))},u=r=>{[E,w,m].forEach(t=>{t&&(t.hidden=t!==r)})},F=(r,t,A)=>{let L=t&&r&&(r.kind==="not_found"||r.status===410);f=r!=null&&r.retryable?A:null,u(null),i("error",L?o.invalid_link||"":x(r,k),!0)},y=r=>{e.setAttribute("aria-busy",r?"true":"false"),e.querySelectorAll("button").forEach(t=>{t.disabled=r})},b={root:e,request:p,async confirm(){let r=!1;y(!0),u(null),i("pending",o.confirming||"");try{await c.feature(h),r=!0;let t=await c.post(`${c.pathFor(h)}/confirm`,{token:p.token});return i("confirmed",o.confirmed||""),t}catch(t){return F(t,r,()=>b.confirm()),null}finally{y(!1)}},async unsubscribe(){let r=!1;y(!0),i("pending",o.unsubscribing||"");try{await c.feature(h),r=!0;let t=await c.delete(d);return u(null),i("unsubscribed",o.unsubscribed||""),t}catch(t){return F(t,r,()=>b.unsubscribe()),null}finally{y(!1)}},async load(){let r=!1;y(!0),u(null),i("loading",o.loading||"");try{await c.feature(h),r=!0;let t=await c.get(d),A=t&&t.data&&Array.isArray(t.data.topics)?t.data.topics.map(String):[];return m&&m.querySelectorAll('[name="topics"]').forEach(L=>{L.checked=A.includes(L.value)}),u(m),i("loaded"),A}catch(t){return F(t,r,()=>b.load()),null}finally{y(!1)}},async save(){let r=Array.from(m.querySelectorAll('[name="topics"]:checked')).map(t=>t.value);y(!0),i("pending",o.saving||"");try{let t=await c.patch(d,{topics:r});return i("saved",o.saved||""),t}catch(t){return f=t!=null&&t.retryable?()=>b.save():null,i("error",x(t,k),!0),null}finally{y(!1)}}};return e.querySelectorAll("[data-manage-unsubscribe]").forEach(r=>{r.addEventListener("click",()=>b.unsubscribe())}),m&&m.addEventListener("submit",r=>{r.preventDefault(),b.save()}),c.enabled?p?(S&&typeof S.replaceState=="function"&&S.replaceState(null,"",g.pathname),p.action==="confirm"?b.confirm():p.action==="manage"?b.load():(u(w),i("idle")),b):(u(E),i("idle"),b):(u(null),i("disabled",k.disabled||""),b)}function D(e=document){return Array.from(e.querySelectorAll("[data-subscription-manage]")).map(a=>V(a))}function _(){U(document),D(document)}document.readyState==="loading"?document.addEventListener("DOMContentLoaded",_,{once:!0}):_();
+import{a as R,b as O,c as T,d as q,e as I,f as U}from"./chunks/chunk-TNVD6UAM.js";import{a as x,c as F}from"./chunks/chunk-VZ5WKQVA.js";import"./chunks/chunk-PATLC23F.js";var h="subscriptions",N=/^[^\s@]+@[^\s@]+\.[^\s@]+$/;function $(e,a){let s=e.querySelector(a);if(!s)return{};try{return JSON.parse(s.textContent)||{}}catch(n){return{}}}function j(e,a={}){let s={email:e.email||""};return a.hasTopics&&(s.topics=[].concat(e.topics||[]).filter(Boolean)),a.sourceUrl&&(s.source_url=a.sourceUrl),a.locale&&(s.locale=a.locale),s}function P(e,a,s={}){let n={};return N.test(e.email)||(n.email=s.email_invalid||"That does not look like an email address."),Array.isArray(e.topics)&&e.topics.length===0&&(n.topics=s.topics_required||"Pick at least one."),a.website&&(n.website="spam"),Object.keys(n).length>0?n:null}function J(e,a={}){let s=a.client||F(),n=e.querySelector("form"),c=$(e,"[data-subscribe-labels]"),g=c.errors||{},S=e.ownerDocument,o=S.querySelector('link[rel="canonical"]'),k={sourceUrl:e.dataset.subscribeSource||o&&o.href||"",locale:e.dataset.subscribeLocale||S.documentElement.lang||"",hasTopics:n.querySelector('[name="topics"]')!==null},w=e.dataset.subscribeDoubleOptIn!=="false",E=R(n),v=d=>{n.dataset.result=d,q(n,"success",c[d]||"")};if(!s.enabled)return q(n,"disabled",g.disabled||""),{form:n,available:!1,submit:async()=>null};let m=O(s,h,n,g);T(n,()=>m().catch(()=>{}));let p={form:n,available:!0,async submit(){let d=I(n),f=j(d,k),l=P(f,d,c);if(delete n.dataset.result,l)return l.website?(v("pending"),null):(q(n,"invalid",g.invalid||"",l),null);q(n,"pending",c.sending||"");try{await m()}catch(i){return null}try{let i=await s.post(s.pathFor(h),f,{idempotencyKey:E.key(f)}),u=i&&i.data&&typeof i.data.status=="string"?i.data.status:"";return E.clear(),n.reset(),v(u==="confirmed"||u==="subscribed"||u===""&&!w?"confirmed":"pending"),i}catch(i){return i&&i.kind==="conflict"?v("duplicate"):U(n,i,g),null}}};return n.addEventListener("submit",d=>{d.preventDefault(),p.submit()}),p}function D(e=document){return Array.from(e.querySelectorAll("[data-subscribe]")).map(a=>J(a))}var K=["confirm","unsubscribe","manage"],z=/^[A-Za-z0-9._~-]{8,512}$/;function B(e,a){let s=e.querySelector(a);if(!s)return{};try{return JSON.parse(s.textContent)||{}}catch(n){return{}}}function V(e){let a=new URLSearchParams(e||""),s=K.find(c=>a.has(c));if(!s)return null;let n=(a.get(s)||"").trim();return z.test(n)?{action:s,token:n}:null}function Z(e,a={}){let s=e.ownerDocument,n=s.defaultView,c=a.client||F(),g=a.location||n&&n.location||{search:"",pathname:"/"},S=a.history===void 0?n&&n.history:a.history,o=B(e,"[data-manage-labels]"),k=o.errors||{},w=e.querySelector("[data-manage-status]"),E=e.querySelector("[data-manage-idle]"),v=e.querySelector("[data-manage-unsubscribe-panel]"),m=e.querySelector("[data-manage-preferences]"),p=V(g.search),d=p?`${c.pathFor(h)}/${encodeURIComponent(p.token)}`:"",f=null,l=s.createElement("button");l.type="button",l.className="post-tool",l.dataset.manageRetry="",l.textContent=o.retry||"Try again",l.hidden=!0,l.addEventListener("click",()=>f==null?void 0:f()),e.appendChild(l);let i=(r,t="",A=!1)=>{r!=="error"&&(f=null),e.dataset.state=r,l.hidden=r!=="error"||!f,w&&(w.textContent=t,w.hidden=t==="",w.setAttribute("role",A?"alert":"status"))},u=r=>{[E,v,m].forEach(t=>{t&&(t.hidden=t!==r)})},L=(r,t,A)=>{let C=t&&r&&(r.kind==="not_found"||r.status===410);f=r!=null&&r.retryable?A:null,u(null),i("error",C?o.invalid_link||"":x(r,k),!0)},y=r=>{e.setAttribute("aria-busy",r?"true":"false"),e.querySelectorAll("button").forEach(t=>{t.disabled=r})},b={root:e,request:p,async confirm(){let r=!1;y(!0),u(null),i("pending",o.confirming||"");try{await c.feature(h),r=!0;let t=await c.post(`${c.pathFor(h)}/confirm`,{token:p.token});return i("confirmed",o.confirmed||""),t}catch(t){return L(t,r,()=>b.confirm()),null}finally{y(!1)}},async unsubscribe(){let r=!1;y(!0),i("pending",o.unsubscribing||"");try{await c.feature(h),r=!0;let t=await c.delete(d);return u(null),i("unsubscribed",o.unsubscribed||""),t}catch(t){return L(t,r,()=>b.unsubscribe()),null}finally{y(!1)}},async load(){let r=!1;y(!0),u(null),i("loading",o.loading||"");try{await c.feature(h),r=!0;let t=await c.get(d),A=t&&t.data&&Array.isArray(t.data.topics)?t.data.topics.map(String):[];return m&&m.querySelectorAll('[name="topics"]').forEach(C=>{C.checked=A.includes(C.value)}),u(m),i("loaded"),A}catch(t){return L(t,r,()=>b.load()),null}finally{y(!1)}},async save(){let r=Array.from(m.querySelectorAll('[name="topics"]:checked')).map(t=>t.value);y(!0),i("pending",o.saving||"");try{let t=await c.patch(d,{topics:r});return i("saved",o.saved||""),t}catch(t){return f=t!=null&&t.retryable?()=>b.save():null,i("error",x(t,k),!0),null}finally{y(!1)}}};return e.querySelectorAll("[data-manage-unsubscribe]").forEach(r=>{r.addEventListener("click",()=>b.unsubscribe())}),m&&m.addEventListener("submit",r=>{r.preventDefault(),b.save()}),c.enabled?p?(S&&typeof S.replaceState=="function"&&S.replaceState(null,"",g.pathname),p.action==="confirm"?b.confirm():p.action==="manage"?b.load():(u(v),i("idle")),b):(u(E),i("idle"),b):(u(null),i("disabled",k.disabled||""),b)}function _(e=document){return Array.from(e.querySelectorAll("[data-subscription-manage]")).map(a=>Z(a))}function M(){D(document),_(document)}document.readyState==="loading"?document.addEventListener("DOMContentLoaded",M,{once:!0}):M();

data/assets/js/loader.js

@@ -74,8 +74,12 @@ const FEATURE_CONFIG = [
     test: () => document.body?.dataset.featureMath === "true"
   },
   {
+    // The citation hooks, and the academic dashboard's submission and calendar
+    // filters, which a dashboard without citation metrics still has.
     name: "academic",
-    test: () => document.querySelector("[data-citation-metric], [data-citation-table], [data-citation-chart]")
+    test: () => document.querySelector(
+      "[data-citation-metric], [data-citation-table], [data-citation-chart], [data-submission-filter], [data-calendar-filter]"
+    )
   },
   {
     name: "notebook",
@@ -88,7 +92,7 @@ const FEATURE_CONFIG = [
   },
   {
     name: "corrections",
-    test: () => document.querySelector("[data-correction-report]")
+    test: () => document.querySelector("[data-correction-report], [data-correction-fallback]")
   },
   {
     name: "contact",

data/lib/datalog/audit/checks.rb

@@ -0,0 +1,269 @@
+# frozen_string_literal: true
+
+require "date"
+require "time"
+require "uri"
+
+module Datalog
+  class Audit
+    # What each check looks for in one file. A check returns [line, message]
+    # pairs; Audit names the check, its kind and the file.
+    class Checks
+      KINDS = %w[theorem lemma proposition corollary definition assumption example remark].freeze
+      # **Definition 2.**, __Theorem (Bayes)__, > **Lemma**, *Proof.*
+      STATEMENT = /\A\s*(?:>\s*)*(?:[-*+]\s+)?(\*\*|__)(#{(KINDS + ['proof']).join('|')})\b[^*_\n]*\1/i
+      PROOF = /\A\s*(?:>\s*)*(\*|_)proof\.?\1/i
+      TYPED_NUMBER = /\b(Figure|Fig\.|Table)\s+(\d+)\b/
+      REFERENCES = /\A\s{0,3}(?:\#{1,6}\s+|\*\*)
+                    (References|Bibliography|Works\ cited|Literature\ cited|Sources)(?:\*\*)?[\s\#]*\z/ix
+      # Addresses are found with one plain pattern and judged by their host,
+      # compared whole: a pattern for the host would also match it inside
+      # another address.
+      URL = %r{https?://[^\s)<>\]"'`]+}i
+      REPOSITORY_HOSTS = %w[github.com gitlab.com bitbucket.org codeberg.org].freeze
+      NOTEBOOK_HOSTS = %w[colab.research.google.com mybinder.org nbviewer.org nbviewer.jupyter.org].freeze
+      DOI_HOSTS = %w[doi.org dx.doi.org].freeze
+      NOTEBOOK_FILE = /\.ipynb\b/i
+      BARE_DOI = %r{\bdoi:\s*10\.\d{4,9}/}i
+      # Link and image text stops at a bracket, so every [ starts a scan that
+      # ends at the next one, and a line of brackets takes linear time.
+      MARKDOWN_IMAGE = /!\[([^\[\]\n]*)\]\(\s*<?([^)\s>]+)/
+      HTML_IMAGE = /<img\b[^>]*>/i
+      IMAGE_FILE = /\.(?:png|jpe?g|gif|svg|webp|avif|bmp|tiff?)\z/i
+      MARKDOWN_LINK = %r{(?<!!)\[[^\[\]\n]*\]\(\s*<?(/(?!/)[^)\s>]*)}
+      HTML_LINK = %r{\bhref\s*=\s*["'](/(?!/)[^"']*)["']}i
+      PART = /\bpart[\s_-]*(\d+|[ivx]+)\b/i
+      # A link to a page of the site, and link text that names another part.
+      SITE_LINK = %r{\[([^\[\]\n]*)\]\(\s*<?/}
+      SEQUENCE_TEXT = /\b(?:part\s+(?:\d+|[ivx]+)|(?:next|previous)\s+(?:part|post|article|chapter))\b/i
+
+      def initialize(reader:, known_keys:, settings:)
+        @reader = reader
+        @known_keys = known_keys
+        @settings = settings
+      end
+
+      # --------------------------------------------------------- opportunities
+
+      def statements(file)
+        file.each_line.filter_map do |line, number|
+          match = line.match(STATEMENT) || line.match(PROOF)
+          next unless match
+
+          kind = (match[2] || "proof").downcase
+          [number, "#{kind.capitalize} typed by hand (#{match[0].strip}): {% #{kind} %} numbers it and " \
+                   "{% ref %} links to it"]
+        end
+      end
+
+      def figures(file)
+        file.each_line.filter_map do |line, number|
+          match = line.match(TYPED_NUMBER)
+          next unless match
+
+          tag = match[1] == "Table" ? "table" : "figure"
+          [number, "\"#{match[0]}\" typed by hand: {% #{tag} %} numbers it and {% ref %} keeps the " \
+                   "number right when the order changes"]
+        end
+      end
+
+      def references(file)
+        return [] if file.body.include?("{% cite") || file.body.include?("{%- cite")
+
+        file.each_line.filter_map do |line, number|
+          next unless line.match?(REFERENCES)
+
+          [number, "a references section written by hand: {% cite %} from a BibTeX file numbers the " \
+                   "citations and lists the works they cite"]
+        end.first(1)
+      end
+
+      def reproducibility(file)
+        return [] if !file.post? || file.front_matter.key?("reproducibility")
+
+        file.each_line do |line, number|
+          found = research_link(line)
+          if found
+            return [[number, "links #{found} but has no reproducibility: block, which sets out the code, " \
+                             "data and environment behind the article"]]
+          end
+        end
+        []
+      end
+
+      def revisions(file)
+        data = file.front_matter
+        return [] unless file.post? && !data.key?("revisions")
+
+        published = to_time(data["date"]) || date_from_name(file.relative)
+        return [] unless published
+
+        key = %w[last_modified_at updated].find { |name| data[name] }
+        edited = key ? to_time(data[key]) : @reader.git_dates[file.relative]
+        return [] unless edited && edited - published > revision_gap
+
+        source = key ? "`#{key}`" : "its last commit"
+        [[key ? file.key_line(key) : 1,
+          "edited #{(edited.to_date - published.to_date).to_i} days after publication (#{source}) with no " \
+          "revisions: entry, which tells readers what changed"]]
+      end
+
+      # Posts whose titles or file names say "Part 2", or that link to the next
+      # part by hand, and are in no series.
+      def series(file)
+        return [] if file.front_matter.key?("series") || !file.post?
+
+        found = []
+        title = file.front_matter["title"].to_s
+        part = title[PART] || File.basename(file.relative)[PART]
+        if part
+          where = title[PART] ? "title" : "file name"
+          found << [file.key_line("title"), "\"#{part}\" in the #{where} and no series: series: numbers the " \
+                                            "parts and links each to the next"]
+        end
+        file.each_line do |line, number|
+          link = line.scan(SITE_LINK).flatten.find { |text| text.match?(SEQUENCE_TEXT) }
+          found << [number, "links to \"#{link}\" by hand: series: links the parts in order"] if link
+        end
+        found.first(2)
+      end
+
+      # -------------------------------------------------------------- problems
+
+      # Fields other themes read and DataLog does not, named so the finding
+      # says where a key came from.
+      FOREIGN = {
+        "Minimal Mistakes" => %w[author_profile seo_type read_time share related sidebar toc_sticky toc_label
+                                 toc_icon header_image],
+        "Just the Docs" => %w[nav_exclude nav_order parent grand_parent has_children has_toc search_exclude],
+        "Chirpy" => %w[pin math_rendering mermaid media_subpath]
+      }.freeze
+
+      def front_matter(file)
+        file.front_matter.keys.reject { |key| @known_keys.include?(key.to_s) }.map do |key|
+          theme = FOREIGN.find { |_, keys| keys.include?(key.to_s) }&.first
+          reason = theme ? "a #{theme} field DataLog ignores" : "a typo, or a field of another theme"
+          [file.key_line(key), "front matter key \"#{key}\" is read by no layout, include or plugin: #{reason} " \
+                               "(audit.known_keys lists a site's own)"]
+        end
+      end
+
+      def images(file)
+        found = []
+        file.each_line do |line, number|
+          line.scan(MARKDOWN_IMAGE) do |alt, src|
+            problem = alt_problem(alt, src)
+            found << [number, "image #{src} #{problem}"] if problem
+          end
+          line.scan(HTML_IMAGE) do
+            tag = Regexp.last_match(0)
+            src = tag[/\bsrc\s*=\s*["']([^"']*)["']/i, 1].to_s
+            alt = tag[/\balt\s*=\s*["']([^"']*)["']/i, 1]
+            problem = alt.nil? ? "has no alt attribute" : (alt_problem(alt, src) unless alt.empty?)
+            found << [number, "image #{src} #{problem}"] if problem
+          end
+        end
+        found
+      end
+
+      def links(file)
+        found = []
+        file.each_line do |line, number|
+          (line.scan(MARKDOWN_LINK) + line.scan(HTML_LINK)).flatten.each do |target|
+            path = strip_baseurl(target)
+            next if path.include?("{{") || ignored_link?(path) || @reader.built?(path)
+
+            found << [number, "links #{target}, which the site does not build"]
+          end
+        end
+        found
+      end
+
+      # Posts only: a page such as the search page loads the engine for math
+      # its script fetches, which the page's own source cannot show.
+      def math(file)
+        data = file.front_matter
+        return [] unless file.post?
+
+        key = data.key?("math") ? "math" : ("mathjax" if data.key?("mathjax"))
+        return [] unless key && [true, false].include?(data[key])
+
+        found = detect_math(file.body)
+        if data[key] == true && !found
+          [[file.key_line(key), "#{key}: true loads the math engine, but the page has no math it can find"]]
+        elsif data[key] == false && found
+          [[file.key_line(key), "#{key}: false leaves the page's math (#{found}) as LaTeX source"]]
+        else
+          []
+        end
+      end
+
+      private
+
+      # The first address on the line that is a repository, a notebook or a DOI,
+      # or a notebook file or DOI written without an address.
+      def research_link(line)
+        line.scan(URL).find { |url| research_address?(url) } ||
+          line.split.find { |word| word.match?(NOTEBOOK_FILE) } || line[BARE_DOI]
+      end
+
+      def research_address?(url)
+        uri = URI.parse(url)
+        host = uri.host.to_s.downcase.delete_prefix("www.")
+        segments = uri.path.to_s.split("/").reject(&:empty?)
+        return segments.size >= 2 if REPOSITORY_HOSTS.include?(host)
+        return segments.first.to_s.start_with?("10.") if DOI_HOSTS.include?(host)
+
+        NOTEBOOK_HOSTS.include?(host) || uri.path.to_s.match?(NOTEBOOK_FILE)
+      rescue URI::InvalidURIError
+        false
+      end
+
+      def alt_problem(alt, src)
+        text = alt.to_s.strip
+        return "has no alt text" if text.empty?
+
+        name = File.basename(src.to_s.split(/[?#]/).first.to_s)
+        "has its file name as alt text" if text.match?(IMAGE_FILE) || text == name || text == File.basename(name, ".*")
+      end
+
+      def strip_baseurl(target)
+        base = @reader.config["baseurl"].to_s.chomp("/")
+        base.empty? || !target.start_with?("#{base}/") ? target : target.delete_prefix(base)
+      end
+
+      def ignored_link?(path)
+        Array(@settings["ignore_links"]).any? { |prefix| path.start_with?(prefix.to_s) }
+      end
+
+      # The first expression the math preprocessor would render, or nil.
+      def detect_math(body)
+        return nil unless defined?(::MathPreprocessor::Processor)
+
+        processor = ::MathPreprocessor::Processor.new(body)
+        processor.process
+        expression = processor.expressions.first
+        expression && expression["latex"].to_s.strip[0, 40]
+      end
+
+      def revision_gap
+        (@settings["revision_after_days"] || 30).to_i * 86_400
+      end
+
+      def to_time(value)
+        case value
+        when Time then value
+        when Date then value.to_time
+        when String then Time.parse(value)
+        end
+      rescue ArgumentError
+        nil
+      end
+
+      def date_from_name(relative)
+        stamp = File.basename(relative)[/\A\d{4}-\d{2}-\d{2}/]
+        stamp && Time.parse(stamp)
+      end
+    end
+  end
+end

data/lib/datalog/audit/known_keys.rb

@@ -0,0 +1,61 @@
+# frozen_string_literal: true
+
+module Datalog
+  class Audit
+    # The front matter keys something reads, found in the sources that read
+    # them rather than kept as a list: the theme's layouts, includes, plugins
+    # and library, and the site's own layouts, includes and plugins. A key
+    # none of them names is a typo (`descripton`) or a leftover of another
+    # theme, which nothing will ever show.
+    module KnownKeys
+      module_function
+
+      # Keys that Jekyll and the plugins the theme depends on read, which the
+      # theme's own sources do not name.
+      JEKYLL = %w[
+        layout permalink published date categories category tags tag title excerpt excerpt_separator slug
+        lang render_with_liquid collection output path url id ext draft hidden
+        sitemap redirect_from redirect_to last_modified_at image seo canonical_url locale author description
+        paginate pagination feed
+      ].freeze
+
+      TEMPLATE_DIRS = %w[_layouts _includes].freeze
+      CODE_DIRS = %w[_plugins lib].freeze
+      # page.title, include.page.summary, post.series and every other property
+      # a template reads, and quoted names in filters (where: "layout").
+      TEMPLATE_KEY = /[\w\]]\.([a-z_][a-z0-9_]*)|["']([a-z_][a-z0-9_]*)["']/
+      # data["reproducibility"], fetch("title"), Authors.value(page, "series").
+      CODE_KEY = /["']([a-z_][a-z0-9_]*)["']/
+
+      def build(theme_root:, site_root:, extra: [])
+        keys = Set.new(JEKYLL)
+        [theme_root, site_root].uniq.each do |root|
+          scan(root, TEMPLATE_DIRS, "**/*.{html,liquid,xml,json,md}", TEMPLATE_KEY, keys)
+          scan(root, CODE_DIRS, "**/*.rb", CODE_KEY, keys)
+        end
+        keys.merge(Array(extra).map(&:to_s))
+      end
+
+      def scan(root, dirs, glob, pattern, keys)
+        dirs.each do |dir|
+          Dir.glob(File.join(root, dir, glob)).each do |file|
+            File.read(file, encoding: "utf-8").scan(pattern) { |match| keys.merge(Array(match).compact) }
+          end
+        end
+      end
+
+      # Liquid output and tags; neither holds a brace, so a scan from one {{
+      # stops at the next brace and the file is read in linear time.
+      LIQUID = /\{\{[^{}]*\}\}|\{%(?:[^{}%]|%(?!\}))*%\}/
+
+      # The properties a page's own Liquid reads, such as a showcase page that
+      # passes `page.academic_demo` to an include, or a listing that shows
+      # `post.subtitle`: read by the site's content rather than its templates.
+      def from_content(text)
+        text.scan(LIQUID).each_with_object(Set.new) do |liquid, keys|
+          liquid.scan(TEMPLATE_KEY) { |match| keys.merge(Array(match).compact) }
+        end
+      end
+    end
+  end
+end

data/lib/datalog/audit/site_reader.rb

@@ -0,0 +1,118 @@
+# frozen_string_literal: true
+
+require "open3"
+require "time"
+require "tmpdir"
+
+module Datalog
+  class Audit
+    # What Jekyll would build, found without building it: the site's own
+    # content files and every URL the build has, read with Jekyll's reader and
+    # the generators that only add pages in memory. Nothing is rendered or
+    # written; the destination is a directory that never exists.
+    class SiteReader
+      # The generators that add pages (feeds, the sitemap, pagination pages,
+      # redirects, the search page, notebook pages) and write nothing until the
+      # site is written. The others are left out: they fetch data, encode
+      # images or fill caches.
+      PAGE_GENERATORS = %w[
+        JekyllFeed::Generator Jekyll::JekyllSitemap Jekyll::Paginate::Pagination
+        JekyllRedirectFrom::Generator Datalog::SearchPages Jekyll::NotebookConverter
+      ].freeze
+      CONTENT = /\.(md|markdown|html?)\z/i
+
+      attr_reader :root, :site
+
+      def initialize(root)
+        @root = File.expand_path(root)
+      end
+
+      def read
+        require "jekyll"
+        Jekyll.logger.log_level = :error
+        config = Jekyll.configuration(
+          "source" => root, "destination" => File.join(Dir.tmpdir, "datalog-audit-#{Process.pid}", "_site"),
+          "quiet" => true, "disable_disk_cache" => true
+        )
+        @site = Jekyll::Site.new(config)
+        @site.reset
+        @site.read
+        @site.generators.each do |generator|
+          generator.generate(@site) if PAGE_GENERATORS.include?(generator.class.name)
+        end
+        self
+      end
+
+      def config
+        site.config
+      end
+
+      # [absolute path, path from the site root] of every page and document the
+      # site's own sources hold, in a stable order.
+      def content_files
+        documents = site.collections.values.flat_map(&:docs)
+        pages = site.pages.select { |page| page.instance_of?(Jekyll::Page) }
+        files = (documents + pages).filter_map do |item|
+          path = File.expand_path(item.relative_path, root)
+          next unless path.start_with?("#{root}/") && File.file?(path) && path.match?(CONTENT)
+
+          [path, path.delete_prefix("#{root}/")]
+        end
+        files.uniq.sort_by(&:last)
+      end
+
+      # Every URL the build writes, normalised: pages, documents, generated
+      # pages and static files.
+      def urls
+        @urls ||= begin
+          items = site.pages + site.static_files + site.collections.values.flat_map(&:docs).select(&:write?)
+          items.to_set { |item| normalize(item.url) }
+        end
+      end
+
+      def normalize(url)
+        path = url.to_s.split(/[?#]/, 2).first.to_s
+        path = "/#{path}" unless path.start_with?("/")
+        path = path.sub(%r{/index\.html?\z}, "/")
+        path.end_with?("/") || path.length == 1 ? path : path.sub(/\.html?\z/, "")
+      end
+
+      def built?(url)
+        path = normalize(url)
+        [path, "#{path}/", path.delete_suffix("/")].any? { |candidate| urls.include?(candidate) }
+      end
+
+      # A commit that touches more files than this is a sweeping change (a
+      # reformat, a migration, a rename), not an edit of any one article.
+      SWEEPING_COMMIT = 10
+
+      # The date of the last commit to edit each file, from one `git log` over
+      # the site, or {} outside a repository.
+      def git_dates
+        @git_dates ||= begin
+          output, status = Open3.capture2("git", "-C", root, "log", "--format=@@date@@%cI", "--name-only", "--relative",
+                                          "--", ".", err: File::NULL)
+          status.success? ? parse_log(output) : {}
+        rescue SystemCallError
+          {}
+        end
+      end
+
+      private
+
+      def parse_log(output)
+        dates = {}
+        output.split("@@date@@").each do |commit|
+          stamp, *paths = commit.split("\n").map(&:strip).reject(&:empty?)
+          next if stamp.nil? || paths.size > SWEEPING_COMMIT
+
+          date = Time.iso8601(stamp)
+          paths.each { |path| dates[path] ||= date }
+        rescue ArgumentError
+          next
+        end
+        dates
+      end
+    end
+  end
+end

data/lib/datalog/audit/source_file.rb

@@ -0,0 +1,85 @@
+# frozen_string_literal: true
+
+require "date"
+require "yaml"
+
+module Datalog
+  class Audit
+    # One content file as its author wrote it: the front matter keys with the
+    # line each is on, and the body with its code set aside, line by line, so
+    # a finding names the line the author would edit.
+    class SourceFile
+      # Code shows what it shows: nothing inside it is a statement, a figure
+      # reference, a link or an image. Each match is blanked character for
+      # character, newlines kept, so line numbers do not move.
+      CODE = [
+        /^[ \t]*(`{3,}|~{3,})[^\n]*\n.*?(?:^[ \t]*\1[ \t]*$|\z)/m,
+        /\{%-?\s*(highlight|raw)\b.*?\{%-?\s*end\1\s*-?%\}/m,
+        %r{<(pre|code)\b[^>]*>.*?</\1>}mi,
+        /`[^`\n]+`/,
+        /<!--.*?-->/m,
+        /\{%-?\s*comment\s*-?%\}.*?\{%-?\s*endcomment\s*-?%\}/m
+      ].freeze
+
+      attr_reader :path, :relative, :front_matter, :key_lines, :body_start, :body, :lines
+
+      def initialize(path, relative)
+        @path = path
+        @relative = relative
+        text = File.read(path, encoding: "bom|utf-8")
+        @front_matter, @key_lines, @body_start, @body = split(text)
+        @lines = mask(@body).split("\n", -1)
+      end
+
+      # The body's lines with code blanked, each with its line in the file.
+      def each_line
+        return enum_for(:each_line) unless block_given?
+
+        @lines.each_with_index { |line, index| yield line, @body_start + index }
+      end
+
+      # The line in the file of the body's `offset`th character.
+      def line_at(offset)
+        @body_start + masked_body[0, offset].count("\n")
+      end
+
+      def masked_body
+        @masked_body ||= @lines.join("\n")
+      end
+
+      def key_line(key)
+        @key_lines.fetch(key, 1)
+      end
+
+      def post?
+        relative.match?(%r{(\A|/)_posts/})
+      end
+
+      private
+
+      def split(text)
+        match = text.match(/\A---[ \t]*\r?\n(.*?)^---[ \t]*\r?$\n?/m)
+        return [{}, {}, 1, text] unless match
+
+        yaml = match[1]
+        data = begin
+          YAML.safe_load(yaml, permitted_classes: [Date, Time], aliases: true)
+        rescue Psych::Exception
+          nil
+        end
+        key_lines = {}
+        yaml.each_line.with_index(2) do |line, number|
+          key = line[/\A([A-Za-z_][\w-]*)\s*:/, 1]
+          key_lines[key] ||= number if key
+        end
+        [data.is_a?(Hash) ? data : {}, key_lines, match[0].count("\n") + 1, match.post_match]
+      end
+
+      def mask(text)
+        CODE.reduce(text) do |masked, pattern|
+          masked.gsub(pattern) { |code| code.gsub(/[^\n]/, " ") }
+        end
+      end
+    end
+  end
+end