implemented a lot of features

lib/subject/file_key.ts

@@ -0,0 +1,30 @@
+import { createHash } from 'node:crypto'
+
+/**
+ * Identity of one uploaded file: every segment of it has the same poster, the
+ * same subject apart from the segment counter, and the same segment count.
+ * The group is deliberately left out so crossposts resolve to one file.
+ * See CONCEPT.md (D12).
+ */
+export function fileKey(poster: string, base: string, totalSegments: number): string {
+  return createHash('sha1').update(`${poster}\n${base}\n${totalSegments}`).digest('hex')
+}
+
+/** True when a yEnc name looks like a real filename (it has an extension), not a random token. */
+export function looksLikeFilename(name: string): boolean {
+  return /\.[A-Za-z0-9]{1,5}$/.test(name) && !/^[0-9a-f]{32,}$/i.test(name)
+}
+
+/**
+ * Identity of a file known only from its yEnc header (obfuscated posts).
+ * The poster is left out: obfuscating posters randomise it per article.
+ * - A real filename is stable across the articles of a file, and needed to
+ *   tell apart the equal-sized volumes of one archive.
+ * - A random, extensionless name can change per article too; the exact
+ *   byte size and part count are what stays the same.
+ * Collisions are caught by the conflict check in IngestService.
+ */
+export function weakFileKey(name: string, size: number, totalParts: number): string {
+  const identity = looksLikeFilename(name) ? `name\n${name}` : 'shape'
+  return createHash('sha1').update(`weak\n${identity}\n${size}\n${totalParts}`).digest('hex')
+}

lib/subject/index.ts

@@ -0,0 +1,5 @@
+/**
+ * Conventions of binary posts (subject lines, file identity), independent of AdonisJS.
+ */
+export { parseSubject, type ParsedSubject } from './parse.js'
+export { fileKey, weakFileKey, looksLikeFilename } from './file_key.js'

lib/subject/parse.ts

@@ -0,0 +1,59 @@
+export interface ParsedSubject {
+  /** Filename from the quoted part of the subject. */
+  name: string
+  /** Segment number `k` from `(k/N)`. */
+  segment: number
+  /** Segment count `N` from `(k/N)`. */
+  totalSegments: number
+  /** File counter `[i/M]`, when present. */
+  fileIndex?: number
+  fileTotal?: number
+  /**
+   * The subject with the segment counter removed. Identical for every segment
+   * of one uploaded file, so it is part of the file identity (see fileKey).
+   */
+  base: string
+}
+
+const SEGMENT_COUNTER = /\((\d+)\/(\d+)\)/g
+const FILE_COUNTER = /\[(\d+)\/(\d+)\]/
+
+/**
+ * Parses a binary post subject such as
+ * `[07/45] - "Some.Release.part06.rar" yEnc (12/137) 51315352`.
+ *
+ * Returns null for subjects without a quoted filename and segment counter
+ * (obfuscated posts), which need their yEnc header read from the body instead.
+ */
+export function parseSubject(subject: string): ParsedSubject | null {
+  const nameEnd = subject.lastIndexOf('"')
+  const nameStart = nameEnd > 0 ? subject.lastIndexOf('"', nameEnd - 1) : -1
+  if (nameStart === -1) return null
+  const name = subject.slice(nameStart + 1, nameEnd)
+  if (!name) return null
+
+  // The segment counter is the last "(k/N)" after the filename; a file size may follow it.
+  let counter: RegExpExecArray | undefined
+  for (const match of subject.matchAll(SEGMENT_COUNTER)) {
+    if (match.index > nameEnd) counter = match
+  }
+  if (!counter) return null
+
+  const segment = Number.parseInt(counter[1], 10)
+  const totalSegments = Number.parseInt(counter[2], 10)
+  if (totalSegments < 1 || segment < 1 || segment > totalSegments) return null
+
+  const base = (subject.slice(0, counter.index) + subject.slice(counter.index + counter[0].length))
+    .replace(/\s+/g, ' ')
+    .trim()
+
+  const fileCounter = FILE_COUNTER.exec(subject.slice(0, nameStart))
+  return {
+    name,
+    segment,
+    totalSegments,
+    fileIndex: fileCounter ? Number.parseInt(fileCounter[1], 10) : undefined,
+    fileTotal: fileCounter ? Number.parseInt(fileCounter[2], 10) : undefined,
+    base,
+  }
+}