code wiki / (root) / nx_fetch.nx

nx_fetch.nx source

↩ module page · 221 lines · 10643 B

1// nx_fetch.nx -- THE sovereign manifest-driven research fetcher. MERGES the N copy-pasted one-time 2// nx_*_research_fetch organs into ONE data-driven tool (rule 11 data-driven + rule 15 DRY): new research is a 3// MANIFEST ROW, never a new organ. Reads a TSV manifest (each line: url<TAB>name), and for each row: 4// 1. ensures the raw archive knowledge/fetched/<name>.raw (fetch over sovereign TLS-1.3 if absent; idempotent), 5// 2. strips it to clean plain text -> knowledge/library/<name>.txt (the dir the librarian nx_library_harvest_v2 6// walks for .txt) so it BANKS into the durable index, queryable via nx_search_cli. 7// Plaintext strip is purpose-fit for the index (drop tags, suppress <script>/<style>, collapse whitespace) -- 8// NOT the browser's nx_html_to_text (which emits GUI link-markers wrong for a corpus). 100% sovereign. 9// usage: nx_fetch <manifest.tsv> expect_exit: 0 license_tier: ORIGINAL 10import "nx_syscalls.nx" 11import "nx_x509_trust_store.nx" 12import "nx_trust_store_load_from_certdata.nx" 13import "nx_https_fetch_follow.nx" 14const K_MAGIC_4096: i64 = 4096 15const K_MAGIC_4194304: i64 = 4194304 16const K_MAGIC_8388608: i64 = 8388608 17const K_MAGIC_1024: i64 = 1024 18 19func puts(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } sys_write(1, s, n); return 0 } 20func putn(v: i64) -> i64 { 21 if v == 0 { sys_write(1, "0" as *u8, 1); return 0 } 22 var m: i64 = v 23 if m < 0 { sys_write(1, "-" as *u8, 1); m = 0 - m } 24 let d: *u8 = sys_mmap(24); var k: i64 = 0 25 while m > 0 { d[k] = (48 + (m % 10)) as u8; m = m / 10; k = k + 1 } 26 var j: i64 = k - 1 27 while j >= 0 { sys_write(1, ((d as i64)+j) as *u8, 1); j = j - 1 } 28 return 0 29} 30func have_file(path: *u8) -> i64 { let fd: i64 = sys_openat_rd(path); if fd < 0 { return 0 } sys_close(fd); return 1 } 31func scat(dst: *u8, o: i64, s: *u8) -> i64 { var i: i64 = 0; while s[i] != (0 as u8) { dst[o + i] = s[i]; i = i + 1 } return o + i } 32func bpath(dst: *u8, dir: *u8, name: *u8, suf: *u8) -> i64 { 33 var o: i64 = scat(dst, 0, dir); o = scat(dst, o, name); o = scat(dst, o, suf); dst[o] = 0 as u8; return o 34} 35func save_file(path: *u8, buf: *u8, n: i64) -> i64 { 36 let fd: i64 = sys_openat_wr(path, 0x1a4) 37 if fd < 0 { return 0 } 38 var off: i64 = 0 39 while off < n { let w: i64 = sys_write(fd, ((buf as i64) + off) as *u8, n - off); if w <= 0 { sys_close(fd); return 0 } off = off + w } 40 sys_close(fd); return 1 41} 42 43// case-insensitive compare src[off..off+litlen) == lit (lit lowercase) 44func tag_eq_ci(src: *u8, off: i64, n: i64, lit: *u8, litlen: i64) -> i64 { 45 if off + litlen > n { return 0 } 46 var i: i64 = 0 47 while i < litlen { 48 var c: i64 = src[off + i] as i64 49 if c >= 0x41 { if c <= 0x5a { c = c + 0x20 } } 50 if c != (lit[i] as i64) { return 0 } 51 i = i + 1 52 } 53 return 1 54} 55 56// tag name at src[off..] equals lit AND is terminated (space/>//tab/nl) -- so "p" won't match <param>. 57func tag_is(src: *u8, off: i64, n: i64, lit: *u8, litlen: i64) -> i64 { 58 if tag_eq_ci(src, off, n, lit, litlen) == 0 { return 0 } 59 let after: i64 = off + litlen 60 if after >= n { return 1 } 61 let c: i64 = src[after] as i64 62 if c == 0x20 { return 1 } 63 if c == 0x3e { return 1 } 64 if c == 0x2f { return 1 } 65 if c == 0x09 { return 1 } 66 if c == 0x0a { return 1 } 67 if c == 0x0d { return 1 } 68 return 0 69} 70 71// is this a BLOCK-level tag? Block boundaries are SENTENCE boundaries -- replacing them with a bare space (the 72// old behaviour) merged e.g. Wikipedia infobox fields into the intro paragraph as one unsplittable run, which 73// broke retrieve-then-read. Emitting a sentence break here is the general readability fix. 74func is_block_tag(src: *u8, off: i64, n: i64) -> i64 { 75 if tag_is(src, off, n, "br" as *u8, 2) == 1 { return 1 } 76 if tag_is(src, off, n, "p" as *u8, 1) == 1 { return 1 } 77 if tag_is(src, off, n, "div" as *u8, 3) == 1 { return 1 } 78 if tag_is(src, off, n, "td" as *u8, 2) == 1 { return 1 } 79 if tag_is(src, off, n, "tr" as *u8, 2) == 1 { return 1 } 80 if tag_is(src, off, n, "th" as *u8, 2) == 1 { return 1 } 81 if tag_is(src, off, n, "li" as *u8, 2) == 1 { return 1 } 82 if tag_is(src, off, n, "ul" as *u8, 2) == 1 { return 1 } 83 if tag_is(src, off, n, "ol" as *u8, 2) == 1 { return 1 } 84 if tag_is(src, off, n, "dd" as *u8, 2) == 1 { return 1 } 85 if tag_is(src, off, n, "dt" as *u8, 2) == 1 { return 1 } 86 if tag_is(src, off, n, "h1" as *u8, 2) == 1 { return 1 } 87 if tag_is(src, off, n, "h2" as *u8, 2) == 1 { return 1 } 88 if tag_is(src, off, n, "h3" as *u8, 2) == 1 { return 1 } 89 if tag_is(src, off, n, "h4" as *u8, 2) == 1 { return 1 } 90 if tag_is(src, off, n, "h5" as *u8, 2) == 1 { return 1 } 91 if tag_is(src, off, n, "h6" as *u8, 2) == 1 { return 1 } 92 if tag_is(src, off, n, "table" as *u8, 5) == 1 { return 1 } 93 if tag_is(src, off, n, "caption" as *u8, 7) == 1 { return 1 } 94 if tag_is(src, off, n, "section" as *u8, 7) == 1 { return 1 } 95 if tag_is(src, off, n, "article" as *u8, 7) == 1 { return 1 } 96 if tag_is(src, off, n, "blockquote" as *u8, 10) == 1 { return 1 } 97 if tag_is(src, off, n, "hr" as *u8, 2) == 1 { return 1 } 98 return 0 99} 100 101// minimal HTML -> plain text for the corpus: drop <...> tags, suppress <script>/<style>, BLOCK tags -> sentence 102// break (readability), collapse ws. 103func strip_html(src: *u8, n: i64, out: *u8, cap: i64) -> i64 { 104 var op: i64 = 0 105 var i: i64 = 0 106 var last_ws: i64 = 1 107 var suppress: i64 = 0 108 while i < n { 109 let c: i64 = src[i] as i64 110 if c == 0x3c { 111 var nameoff: i64 = i + 1 112 var isclose: i64 = 0 113 if nameoff < n { if (src[nameoff] as i64) == 0x2f { isclose = 1; nameoff = nameoff + 1 } } 114 var iss: i64 = 0 115 if tag_eq_ci(src, nameoff, n, "script" as *u8, 6) == 1 { iss = 1 } 116 if tag_eq_ci(src, nameoff, n, "style" as *u8, 5) == 1 { iss = 1 } 117 if iss == 1 { if isclose == 1 { suppress = 0 } else { suppress = 1 } } 118 let isblock: i64 = is_block_tag(src, nameoff, n) 119 i = i + 1 120 var g: i64 = 0 121 while g == 0 { if i >= n { g = 1 } else { if (src[i] as i64) == 0x3e { g = 1 } else { i = i + 1 } } } 122 if i < n { i = i + 1 } 123 if suppress == 0 { 124 if isblock == 1 { 125 // block boundary -> sentence break ". " (dropped as an empty sentence if it lands on ws) 126 if op < cap { if last_ws == 0 { out[op] = 0x20 as u8; op = op + 1 } } 127 if op < cap { out[op] = 0x2e as u8; op = op + 1 } 128 if op < cap { out[op] = 0x20 as u8; op = op + 1 } 129 last_ws = 1 130 } else { 131 if last_ws == 0 { if op < cap { out[op] = 0x20 as u8; op = op + 1; last_ws = 1 } } 132 } 133 } 134 } else { 135 if suppress == 1 { i = i + 1 } 136 else { 137 var isws: i64 = 0 138 if c == 0x20 { isws = 1 } 139 if c == 0x09 { isws = 1 } 140 if c == 0x0a { isws = 1 } 141 if c == 0x0d { isws = 1 } 142 if isws == 1 { 143 if last_ws == 0 { if op < cap { out[op] = 0x20 as u8; op = op + 1; last_ws = 1 } } 144 } else { 145 if op < cap { out[op] = c as u8; op = op + 1; last_ws = 0 } 146 } 147 i = i + 1 148 } 149 } 150 } 151 return op 152} 153 154// ensure raw archive (fetch if absent) -> strip -> library .txt. Returns 1 on a banked row. 155func fetch_to_lib(url: *u8, name: *u8, store: *TrustStore, out: *u8, cap: i64) -> i64 { 156 let rawpath: *u8 = sys_mmap(K_MAGIC_4096); bpath(rawpath, "knowledge/fetched/" as *u8, name, ".raw" as *u8) 157 let txtpath: *u8 = sys_mmap(K_MAGIC_4096); bpath(txtpath, "knowledge/library/" as *u8, name, ".txt" as *u8) 158 let lenbox: *i64 = sys_mmap(16) as *i64 159 var raw: *u8 = 0 as *u8 160 var n: i64 = 0 161 if have_file(rawpath) == 1 { 162 raw = sys_read_file(rawpath, lenbox); n = lenbox[0] 163 puts(name); puts(" [raw cached]") 164 } else { 165 let status: *i64 = sys_mmap(8) as *i64 166 n = nx_https_fetch_follow(url, store, out, cap, 6, status) 167 puts(name); puts(" status="); putn(status[0]) 168 if n <= 0 { puts(" FETCH-FAIL\n"); return 0 } 169 save_file(rawpath, out, n) 170 raw = out 171 } 172 let txt: *u8 = sys_mmap(cap) 173 let tn: i64 = strip_html(raw, n, txt, cap) 174 if save_file(txtpath, txt, tn) == 0 { puts(" TXT-SAVE-FAIL\n"); return 0 } 175 puts(" raw="); putn(n); puts(" -> lib.txt="); putn(tn); puts(" BANKED\n") 176 return 1 177} 178 179func main(argc: i64, argv: *i64) -> i64 { 180 if argc < 2 { puts("usage: nx_fetch <manifest.tsv>\n" as *u8); return 1 } 181 let manifest: *u8 = argv[1] as *u8 182 let r: i64 = nx_trust_store_load_from_certdata("data/mozilla_certdata.txt" as *u8, 512, K_MAGIC_4194304) 183 if r <= 0 { puts("nx_fetch: certdata load failed\n" as *u8); return 1 } 184 let store: *TrustStore = r as *TrustStore 185 sys_mkdir("knowledge/fetched" as *u8, 0x1ed) 186 sys_mkdir("knowledge/library" as *u8, 0x1ed) 187 puts("CA roots="); putn(trust_store_count(store)); puts(" manifest="); puts(manifest); puts("\n") 188 189 let lb: *i64 = sys_mmap(16) as *i64 190 let mdata: *u8 = sys_read_file(manifest, lb) 191 if (mdata as i64) == 0 { puts("nx_fetch: cannot read manifest\n" as *u8); return 1 } 192 let mn: i64 = lb[0] 193 let cap: i64 = K_MAGIC_8388608 194 let out: *u8 = sys_mmap(cap) 195 let urlbuf: *u8 = sys_mmap(K_MAGIC_4096) 196 let namebuf: *u8 = sys_mmap(K_MAGIC_1024) 197 198 var pos: i64 = 0 199 var ok: i64 = 0 200 var nrows: i64 = 0 201 while pos < mn { 202 var eol: i64 = pos 203 var fe: i64 = 0 204 while fe == 0 { if eol >= mn { fe = 1 } else { if (mdata[eol] as i64) == 10 { fe = 1 } else { eol = eol + 1 } } } 205 var tab: i64 = pos 206 var ft: i64 = 0 207 while ft == 0 { if tab >= eol { ft = 1 } else { if (mdata[tab] as i64) == 9 { ft = 1 } else { tab = tab + 1 } } } 208 if tab < eol { 209 var a: i64 = 0 210 while pos + a < tab { urlbuf[a] = mdata[pos + a]; a = a + 1 } 211 urlbuf[a] = 0 as u8 212 var b: i64 = 0 213 while tab + 1 + b < eol { namebuf[b] = mdata[tab + 1 + b]; b = b + 1 } 214 namebuf[b] = 0 as u8 215 if b > 0 { ok = ok + fetch_to_lib(urlbuf, namebuf, store, out, cap); nrows = nrows + 1 } 216 } 217 pos = eol + 1 218 } 219 puts("BANKED-READY: "); putn(ok); puts(" / "); putn(nrows); puts(" rows -> knowledge/library/ (run nx_library_harvest_v2 to index)\n") 220 return 0 221}