code wiki / (root) / nx_media_sniff.nx

nx_media_sniff.nx

buildroot/runtime/nx_media_sniff.nx

52799 B1200 linesdepth 3pulls 3 transitivereach 19 importersview sourcekind librarytopic media

docs dependencies structs consts functions

about

nx_media_sniff.nx -- determine media type from magic bytes / MIME / extension. Four-tier detection (highest trust first): 1. Magic-byte signature (binary formats; reads actual content) 2. MIME-string parsing (when the source provided Content-Type) 3. URL/path extension (text formats + container disambiguation) 4. mostly-printable (LAST-RESORT fallback -> TEXT) Returns NX_MEDIA_TYPE_* (see nx_media_pool.nx). Returns NX_MEDIA_TYPE_UNKNOWN when all four tiers fail. ===== 2026-07-30 REWRITE (rule 3: rewrite, do not patch-cascade) ===== ROOT DEFECT MEASURED: the old sniffer knew only 6 magics (PNG/JPEG/WebP/GIF/ MP4/WebM + MP3/WAV/OGG). It had NO MPEG-TS detector and no ".ts" extension row, so the ENTIRE 53,567-file cam recording corpus migrated into the vault as type=unknown (proven by reading live NXR1 records in knowledge/store/mvault-seg-1785089367.docs: every _cam.ts record showed type=unknown while _cam.jpg posters showed image). MV_TYPE_MODEL3D and MV_TYPE_BOOK were likewise unreachable from bytes -- there was no sniff code to return. This rewrite closes the door properly: STL/PLY/GLB/FBX 3D, PDF/EPUB/MOBI/DJVU books, WARC web captures, and the full A/V container set. TRUST-ORDER FIX: "mostly printable" is a FALLBACK HEURISTIC, not a magic. The old code returned it from inside the magic tier, which meant an ASCII STL or a .obj mesh was sealed as TEXT before the extension tier ever ran. nx_sniff_bytes_strict() omits that fallback so the composite can order the tiers honestly; nx_sniff_bytes() keeps its original contract for existing callers. CONTAINER REFINEMENT: ZIP and generic-TEXT are CONTAINER verdicts, not final ones -- .epub/.cbz/.3mf/.wacz are all "PK\x03\x04", and .obj/.gcode/.dae are all plain text. For those two verdicts ONLY, the extension tier refines the answer. Every other magic still wins outright. Per the boundary cardinal: sniffer is a TRUST BOUNDARY -- treats every byte as adversarial. Always reads through bounded loops per the JPL Rule 2 cardinal. No const arrays (the CONST[i] trap compiles clean then reads garbage); every signature is a literal. See: docs/NISHI_MEDIA_INGESTION_ROADMAP.md

dependencies 2 imports · 8 importers

imports: nx_syscalls.nxnx_media_pool.nx

imported by: nx_entity_media.nxnx_media_fetch.nxnx_media_gather.nxnx_mvault.nxnx_mvault_ingest.nxnx_mvault_reindex.nxnx_mvault_walk.nxnx_text_ingest_test.nx

structs

none

consts

none

functions

72	func nx_sniff_load_u8(p: *u8, i: i64, n: i64) -> i64 called by 30: nx_sniff_at nx_sniff_u32le nx_sniff_is_png nx_sniff_is_jpeg nx_sniff_is_gif nx_sniff_is_webp+24
81	func nx_sniff_at(b: u8, n: i64, off: i64, lit: u8, llen: i64) -> i64 called by 29: nx_sniff_find nx_sniff_is_bmp nx_sniff_is_psd nx_sniff_is_qoi nx_sniff_is_dds nx_sniff_isobmff_kind+23 calls 1: nx_sniff_load_u8
95	func nx_sniff_find(b: u8, n: i64, limit: i64, lit: u8, llen: i64) -> i64 called by 2: nx_sniff_is_stl_ascii nx_sniff_is_svg_token calls 1: nx_sniff_at
108	func nx_sniff_u32le(b: *u8, n: i64, off: i64) -> i64 called by 1: nx_sniff_is_stl_binary calls 1: nx_sniff_load_u8
121	func nx_sniff_is_png(b: *u8, n: i64) -> i64 called by 2: nx_sniff_bytes_strict main calls 1: nx_sniff_load_u8
135	func nx_sniff_is_jpeg(b: *u8, n: i64) -> i64 called by 2: nx_sniff_bytes_strict main calls 1: nx_sniff_load_u8
144	func nx_sniff_is_gif(b: *u8, n: i64) -> i64 called by 2: nx_sniff_bytes_strict main calls 1: nx_sniff_load_u8
159	func nx_sniff_is_webp(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
173	func nx_sniff_is_bmp(b: *u8, n: i64) -> i64 calls 1: nx_sniff_at
180	func nx_sniff_is_tiff(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
200	func nx_sniff_is_ico(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
213	func nx_sniff_is_psd(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
219	func nx_sniff_is_qoi(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
225	func nx_sniff_is_dds(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
231	func nx_sniff_is_exr(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
241	func nx_sniff_is_jxl(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
249	func nx_sniff_is_netpbm(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
267	func nx_sniff_is_mp4(b: *u8, n: i64) -> i64 called by 2: nx_sniff_isobmff_kind nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
279	func nx_sniff_isobmff_kind(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 2: nx_sniff_is_mp4 nx_sniff_at
313	func nx_sniff_is_webm(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
328	func nx_sniff_is_mpegts(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
338	func nx_sniff_is_m2ts(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
347	func nx_sniff_is_mpegps(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
359	func nx_sniff_is_avi(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
367	func nx_sniff_is_flv(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 2: nx_sniff_at nx_sniff_load_u8
375	func nx_sniff_is_asf(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
391	func nx_sniff_is_mp3(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
408	func nx_sniff_is_wav(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
422	func nx_sniff_is_ogg(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
432	func nx_sniff_is_flac(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
438	func nx_sniff_is_aiff(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
449	func nx_sniff_is_midi(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
455	func nx_sniff_is_ape(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
467	func nx_sniff_is_glb(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
473	func nx_sniff_is_ply(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 2: nx_sniff_at nx_sniff_load_u8
485	func nx_sniff_is_stl_ascii(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 2: nx_sniff_at nx_sniff_find
497	func nx_sniff_is_stl_binary(b: *u8, n: i64, filesize: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 2: nx_sniff_at nx_sniff_u32le
510	func nx_sniff_is_fbx(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
516	func nx_sniff_is_blend(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
522	func nx_sniff_is_vox(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
528	func nx_sniff_is_step(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
536	func nx_sniff_is_pdf(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
543	func nx_sniff_is_epub(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
551	func nx_sniff_is_mobi(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
557	func nx_sniff_is_djvu(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
563	func nx_sniff_is_chm(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
569	func nx_sniff_is_rtf(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
582	func nx_sniff_is_warc(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
588	func nx_sniff_is_mhtml(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
598	func nx_sniff_is_zip(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
608	func nx_sniff_is_rar(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_at
614	func nx_sniff_is_7z(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
629	func nx_sniff_is_html_token(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 1: nx_sniff_load_u8
653	func nx_sniff_is_svg_token(b: *u8, n: i64) -> i64 called by 1: nx_sniff_bytes_strict calls 2: nx_sniff_at nx_sniff_find
660	func nx_sniff_is_mostly_printable(b: *u8, n: i64) -> i64 called by 2: nx_sniff_bytes nx_sniff_full calls 1: nx_sniff_load_u8
691	func nx_sniff_bytes_strict(b: *u8, n: i64, filesize: i64) -> i64 called by 2: nx_sniff_bytes nx_sniff_full calls 49: nx_sniff_is_epub nx_sniff_is_pdf nx_sniff_is_mobi nx_sniff_is_djvu nx_sniff_is_chm nx_sniff_is_warc+43
768	func nx_sniff_bytes(b: *u8, n: i64) -> i64 called by 1: main calls 2: nx_sniff_bytes_strict nx_sniff_is_mostly_printable
782	func nx_sniff_mime_eq(mime: u8, mlen: i64, lit: u8, llen: i64) -> i64 called by 1: nx_sniff_mime calls 1: nx_sniff_load_u8
800	func nx_sniff_mime(mime: *u8, mlen: i64) -> i64 called by 2: nx_sniff_full main calls 1: nx_sniff_mime_eq
869	func nx_sniff_ext_eq(p: *u8, plen: i64, off: i64, called by 4: nx_sniff_ext_image nx_sniff_ext_av nx_sniff_ext_model3d nx_sniff_ext_doc calls 1: nx_sniff_load_u8
882	func nx_sniff_ext_off(path: *u8, plen: i64) -> i64 called by 2: nx_sniff_extension do_retype calls 1: nx_sniff_load_u8
903	func nx_sniff_ext_image(p: *u8, plen: i64, off: i64) -> i64 called by 1: nx_sniff_extension calls 1: nx_sniff_ext_eq
959	func nx_sniff_ext_av(p: *u8, plen: i64, off: i64) -> i64 called by 1: nx_sniff_extension calls 1: nx_sniff_ext_eq
1029	func nx_sniff_ext_model3d(p: *u8, plen: i64, off: i64) -> i64 called by 1: nx_sniff_extension calls 1: nx_sniff_ext_eq
1075	func nx_sniff_ext_doc(p: *u8, plen: i64, off: i64) -> i64 called by 1: nx_sniff_extension calls 1: nx_sniff_ext_eq
1147	func nx_sniff_extension(path: *u8, plen: i64) -> i64 called by 3: nx_sniff_full do_retype main calls 5: nx_sniff_ext_off nx_sniff_ext_image nx_sniff_ext_av nx_sniff_ext_model3d nx_sniff_ext_doc
1169	func nx_sniff_full(b: *u8, n: i64, filesize: i64, called by 2: nx_sniff mv_type_of_path calls 4: nx_sniff_bytes_strict nx_sniff_extension nx_sniff_mime nx_sniff_is_mostly_printable
1196	func nx_sniff(b: *u8, n: i64, called by 5: nx_fetch_path_into_pool nx_fetch_stdin_into_pool nx_fetch_register_bytes mv_type_bytes mv_classify_plan calls 1: nx_sniff_full