Content probe: don't download the first segment again for whole files

Par2 sets, NZBs and SFVs were sniffed from their first segment and then
read whole, downloading that segment a second time.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

lib/content/probe.ts

@@ -93,32 +93,29 @@ export function decodeBody(body: Buffer): DecodedSegment {
   }
 }
 
-/** All of a (small) file whose first segment is already decoded; null when it is too big. */
+/**
-export async function readWhole(first: Buffer, segments: TaskSegment[], fetch: BodyFetcher): Promise<Buffer | null> {
+ * All of a (small) file whose first segment is already decoded; null when it is too big.
+ * The first segment isn't downloaded again (it was, for every par2 set, NZB and SFV).
+ */
+export async function readWhole(first: DecodedSegment, segments: TaskSegment[], fetch: BodyFetcher): Promise<Buffer | null> {
   const bytes = segments.reduce((sum, p) => sum + (p.size ?? 0), 0)
   if (segments.length > WHOLE_MAX_PARTS || bytes > WHOLE_MAX_BYTES) return null
-  if (segments.length === 1) return first
+  if (segments.length === 1) return first.data
-  let assembled: YencFile | undefined
+  const assembled = YencFile.fromSegment(first)
-  for (const part of segments) {
+  assembled.add(first)
-    const segment = decodeSegment(await fetch(part.id), { lenientSize: true })
+  for (const part of segments.slice(1)) assembled.add(decodeSegment(await fetch(part.id), { lenientSize: true }))
-    assembled ??= YencFile.fromSegment(segment)
+  return assembled.getBuffer()
-    assembled.add(segment)
-  }
-  return assembled!.getBuffer()
 }
 
 /** A par2 file whose first segment is already decoded, with its other parts if any; null when it is not par2. */
-export async function readPar2(first: Buffer, segments: TaskSegment[], fetch: BodyFetcher): Promise<Par2Set | null> {
+export async function readPar2(first: DecodedSegment, segments: TaskSegment[], fetch: BodyFetcher): Promise<Par2Set | null> {
-  if (!isPar2(first)) return null
+  if (!isPar2(first.data)) return null
   if (segments.length > PAR2_MAX_PARTS) return null
-  if (segments.length === 1) return parsePar2(first)
+  if (segments.length === 1) return parsePar2(first.data)
-  let assembled: YencFile | undefined
+  const assembled = YencFile.fromSegment(first)
-  for (const part of segments) {
+  assembled.add(first)
-    const segment = decodeSegment(await fetch(part.id))
+  for (const part of segments.slice(1)) assembled.add(decodeSegment(await fetch(part.id)))
-    assembled ??= YencFile.fromSegment(segment)
+  return parsePar2(assembled.getBuffer())
-    assembled.add(segment)
-  }
-  return parsePar2(assembled!.getBuffer())
 }
 
 export function isNzbEntry(entry: Pick<RarEntry, 'name' | 'encrypted' | 'directory'>): boolean {
@@ -151,14 +148,14 @@ export async function probeInspection(
     return probe
   }
   if (/\.sfv$/i.test(task.name)) {
-    probe.text = (await readWhole(segment.data, task.segments, fetch))?.toString('latin1') ?? null
+    probe.text = (await readWhole(segment, task.segments, fetch))?.toString('latin1') ?? null
     return probe
   }
   if (contentType === 'nzb') {
-    probe.text = (await readWhole(segment.data, task.segments, fetch))?.toString('utf8') ?? null
+    probe.text = (await readWhole(segment, task.segments, fetch))?.toString('utf8') ?? null
     return probe
   }
-  probe.par2 = await readPar2(segment.data, task.segments, fetch)
+  probe.par2 = await readPar2(segment, task.segments, fetch)
   if (probe.par2) return probe
   if (contentType === '7z') {
     probe.archive = task.sevenZip ? await listSevenZip(segment.data, task.sevenZip.volumes, task.sevenZip.last, fetch, warn) : null
@@ -168,7 +165,7 @@ export async function probeInspection(
   probe.archive = info
   if (info && hasNzbEntries(info)) {
     probe.archiveNzbs = []
-    const archive = await readWhole(segment.data, task.segments, fetch)
+    const archive = await readWhole(segment, task.segments, fetch)
     const archived = archive && parseRarHeaders(archive)
     const entries = archived ? archived.entries.filter(isNzbEntry) : []
     probe.archiveNzbEntries = entries.length

tests/unit/probe.spec.ts

@@ -59,7 +59,8 @@ test.group('content probe', () => {
     const probe = await probeInspection(task, fetch)
     assert.equal(probe.contentType, 'nzb')
     assert.equal(probe.text, xml)
-    assert.lengthOf(fetched, task.segments.length + 1) // the first segment twice: once to sniff, once in the whole
+    // Every segment once: the first, decoded to sniff, is reused for the whole.
+    assert.deepEqual(fetched, task.segments.map((s) => s.id))
   })
 
   test('a 7z archive with encrypted headers is known from its tail, without bsdtar', async ({ assert }) => {