`nx_lib_pdf.nx` source

1// nx_lib_pdf.nx -- sovereign PDF -> text for the library (retires docling).
2// WIRES the ecosystem's nx_pdf_text (objects -> FlateDecode -> per-font CMap ->
3// content-stream Tj/TJ -> UTF-8) + nx_lib_fetch. PROVEN LIVE: fetched the 2.2MB
4// arXiv:1706.03762 PDF over sovereign TLS-1.3 and extracted 536KB of correct
5// text ("1 Introduction Recurrent neural networks ...") -- no poppler/pdfminer/
6// docling, no python. license_tier: ORIGINAL
7// genealogy_id: nishi_library_sovereign_pdf_2026_07_01
8import "nx_lib_fetch.nx"
9import "nx_pdf_text.nx"
10const K_MAGIC_16777216: i64 = 16777216
11
12// extract UTF-8 text from PDF bytes already in memory. returns text length.
13func nx_lib_pdf_text(pdf: *u8, pdflen: i64, out: *u8, out_cap: i64) -> i64 {
14    return nx_pdf_extract_text(pdf, pdflen, out, out_cap)
15}
16
17// fetch a PDF url over sovereign TLS-1.3, strip HTTP headers, extract its text.
18// returns text length, or a negative fetch error.
19func nx_lib_pdf_fetch_text(store: *TrustStore, url: *u8, out: *u8, out_cap: i64) -> i64 {
20    let cap: i64 = K_MAGIC_16777216
21    let buf: *u8 = sys_mmap(cap)
22    let gc: i64 = nx_lib_fetch(store, url, buf, cap)
23    if gc < 0 { return gc }
24    var bs: i64 = 0 - 1
25    var bi: i64 = 0
26    while bi + 3 < gc {
27        if bs < 0 { if buf[bi] == 13 as u8 { if buf[bi+1] == 10 as u8 { if buf[bi+2] == 13 as u8 { if buf[bi+3] == 10 as u8 { bs = bi + 4 } } } } }
28        bi = bi + 1
29    }
30    var bp: *u8 = buf
31    var bl: i64 = gc
32    if bs >= 0 { bp = ((buf as i64) + bs) as *u8; bl = gc - bs }
33    return nx_pdf_extract_text(bp, bl, out, out_cap)
34}
nx_lib_pdf.nx source

`nx_lib_pdf.nx` source
