code wiki / _hdl_build / nx_wiki_exceed.nx
`nx_wiki_exceed.nx` source

↩ module page · 176 lines · 7787 B
1// nx_wiki_exceed.nx -- MEASURED head-to-head of a Nishi wiki page vs the incumbent wiki engine.
2//
3// module: nishi-core.wiki.exceed_census
4// capability: CORE_COMPUTE (the honest "how far is our page from a real wiki page" measurement)
5//
6// Per the no-wave law: an EXCEED is MEASURED per-feature, never self-graded. This organ reads
7// knowledge/registry/wiki_feature_census.tsv (the MediaWiki/Wikipedia feature bar) and SCANS the actual
8// rendered page for each feature's signal -> PRESENT/ABSENT, COMPUTED, not asserted. It reports how many
9// of the incumbent's features our page has, and is HONEST that being behind is the worklist (it refuses
10// to call a behind page an "exceed"). The wiki PAGE BUILDER climbs this census rung by rung.
11// Sovereign: imports only nx_syscalls. license_tier: ORIGINAL
12
13import "nx_syscalls.nx"
14import "nx_itoa_lib.nx"   // shared MSB-first emitter (zero-alloc)
15
16const WE_CENSUS: *u8 = "knowledge/registry/wiki_feature_census.tsv"
17const WE_PAGE:   *u8 = "web_assets/charter.html"
18
19static WE_N:       i64
20static WE_FID:     i64   // *i64 of *u8
21static WE_INC:     i64   // *i64 of i64 (incumbent_has 0/1)
22static WE_SIG:     i64   // *i64 of *u8
23static WE_NOTE:    i64   // *i64 of *u8
24static WE_PRESENT: i64   // *i64 of i64 (computed present 0/1)
25
26func we_puts(s: *u8) -> i64 { var n: i64 = 0; while s[n] != 0 as u8 { n = n + 1 } sys_write(1, s, n); return 0 }
27// MIGRATED to the shared emitter (debt 1785563586). The old body mmapped a scratch buffer
28// per call and never freed it. At PAGE granularity that is 4096B leaked PER CALL -- the
29// defect that took 28.5GB of a 36GB host in nx_ts_lumadiff (2MB input, ~3.66M calls).
30// nxi_* is MSB-first, allocates NOTHING, and emits identical bytes including the sign.
31func we_putn(v: i64) -> i64 { nxi_out(v); return 0 }
32func we_slen(s: *u8) -> i64 { var n: i64 = 0; while s[n] != 0 as u8 { n = n + 1 } return n }
33
34func we_field_dup(b: *u8, ls: i64, le: i64, f: i64) -> *u8 {
  var cur: i64 = 0
  var i: i64 = ls
  while cur < f {
      if i >= le { let e: *u8 = sys_mmap(2); e[0] = 0 as u8; return e }
      if b[i] == 9 as u8 { cur = cur + 1 }
      i = i + 1
  }
  let out: *u8 = sys_mmap(256)
  var o: i64 = 0
  while i < le {
      if b[i] == 9 as u8 { i = le } else { if o < 255 { out[o] = b[i]; o = o + 1 } i = i + 1 }
  }
  out[o] = 0 as u8
  return out
49}
50
51// substring contains: 1 if needle occurs in hay[0..hn). flag-style (no break/continue in nested ifs).
52func we_contains(hay: *u8, hn: i64, needle: *u8) -> i64 {
  let nl: i64 = we_slen(needle)
  if nl == 0 { return 0 }
  if hn < nl { return 0 }
  var i: i64 = 0
  let last: i64 = hn - nl
  var hit: i64 = 0
  while i <= last {
      if hit == 0 {
          var j: i64 = 0; var ok: i64 = 1
          while j < nl { if hay[i + j] != needle[j] { ok = 0; j = nl } else { j = j + 1 } }
          if ok == 1 { hit = 1 }
      }
      i = i + 1
  }
  return hit
68}
69
70// MEASURE: parse the census, scan `page_file`, fill module globals + outs.
71//   outs[0]=n_features outs[1]=incumbent_total outs[2]=ours_match(of incumbent) outs[3]=behind
72func we_measure(census_file: *u8, page_file: *u8, outs: *i64) -> i64 {
  let lenbox: *i64 = sys_mmap(16) as *i64; lenbox[0] = 0
  let data: *u8 = sys_read_file(census_file, lenbox)
  if (data as i64) == 0 { return 0 - 1 }
  let n: i64 = lenbox[0]
  var rows: i64 = 0; var ls: i64 = 0; var i: i64 = 0
  while i <= n {
      var atend: i64 = 0
      if i == n { atend = 1 }
      if i < n { if data[i] == 10 as u8 { atend = 1 } }
      if atend == 1 { if i > ls { if data[ls] != 35 as u8 { rows = rows + 1 } } ls = i + 1 }
      i = i + 1
  }
  let fid: *i64 = sys_mmap(8 * (rows + 1)) as *i64
  let inc: *i64 = sys_mmap(8 * (rows + 1)) as *i64
  let sig: *i64 = sys_mmap(8 * (rows + 1)) as *i64
  let note: *i64 = sys_mmap(8 * (rows + 1)) as *i64
  let present: *i64 = sys_mmap(8 * (rows + 1)) as *i64
  var r: i64 = 0
  ls = 0; i = 0
  while i <= n {
      var atend: i64 = 0
      if i == n { atend = 1 }
      if i < n { if data[i] == 10 as u8 { atend = 1 } }
      if atend == 1 {
          if i > ls { if data[ls] != 35 as u8 {
              fid[r] = we_field_dup(data, ls, i, 0) as i64
              let inctok: *u8 = we_field_dup(data, ls, i, 1)
              var iv: i64 = 0
              if inctok[0] == 49 as u8 { iv = 1 }
              inc[r] = iv
              sig[r] = we_field_dup(data, ls, i, 2) as i64
              note[r] = we_field_dup(data, ls, i, 4) as i64
              present[r] = 0
              r = r + 1
          } }
          ls = i + 1
      }
      i = i + 1
  }
  WE_N = r
  WE_FID = fid as i64; WE_INC = inc as i64; WE_SIG = sig as i64; WE_NOTE = note as i64; WE_PRESENT = present as i64
114
  // read the target page (unreadable -> treated as all-absent, the honest worst case).
  let pbox: *i64 = sys_mmap(16) as *i64; pbox[0] = 0
  let page: *u8 = sys_read_file(page_file, pbox)
  var pn: i64 = 0
  if (page as i64) != 0 { pn = pbox[0] }
120
  var inc_total: i64 = 0; var ours_match: i64 = 0; var behind: i64 = 0; var exceed: i64 = 0
  var c: i64 = 0
  while c < WE_N {
      var pr: i64 = 0
      if pn > 0 { pr = we_contains(page, pn, sig[c] as *u8) }
      present[c] = pr
      if inc[c] == 1 {
          inc_total = inc_total + 1
          if pr == 1 { ours_match = ours_match + 1 } else { behind = behind + 1 }
      }
      if inc[c] == 0 { if pr == 1 { exceed = exceed + 1 } }
      c = c + 1
  }
  outs[0] = WE_N; outs[1] = inc_total; outs[2] = ours_match; outs[3] = behind; outs[4] = exceed
  return 0
136}
137
138func main() -> i64 {
  we_puts("=== NISHI WIKI PAGE vs INCUMBENT (MediaWiki/Wikipedia) -- MEASURED feature census ===\n")
  let outs: *i64 = sys_mmap(256) as *i64
  if we_measure(WE_CENSUS, WE_PAGE, outs) != 0 {
      we_puts("FATAL: "); we_puts(WE_CENSUS); we_puts(" unreadable\n"); sys_exit(1); return 1
  }
  we_puts("  page measured: "); we_puts(WE_PAGE); we_puts("\n")
  let fids: *i64 = WE_FID as *i64
  let incs: *i64 = WE_INC as *i64
  let notes: *i64 = WE_NOTE as *i64
  let pres: *i64 = WE_PRESENT as *i64
  var i: i64 = 0
  while i < WE_N {
      if pres[i] == 1 { we_puts("    [x] ") } else { we_puts("    [ ] ") }
      we_puts(fids[i] as *u8)
      if incs[i] == 1 { we_puts("  (MediaWiki: yes)  ") } else { we_puts("  (MediaWiki: no)   ") }
      we_puts(notes[i] as *u8); we_puts("\n")
      i = i + 1
  }
  we_puts("  MEASURED: ours="); we_putn(outs[2])
  we_puts("/"); we_putn(outs[1])
  we_puts(" standard incumbent features present; BEHIND by "); we_putn(outs[3])
  we_puts("; EXCEED axes present (incumbent lacks)="); we_putn(outs[4]); we_puts(".\n")
  if outs[3] == 0 {
      if outs[4] > 0 {
          we_puts("  VERDICT: MEASURED S-class -- PARITY on all "); we_putn(outs[1])
          we_puts(" standard wiki features AND exceeds on "); we_putn(outs[4])
          we_puts(" axis(es) the incumbent structurally lacks (machine-verified live evidence, no-drift regeneration).\n")
          we_puts("           NARROW + HONEST: the exceed is on integrity/freshness for a self-documenting system -- NOT a claim of beating Wikipedia on content/scale/community.\n")
      } else {
          we_puts("  VERDICT: PARITY on all "); we_putn(outs[1])
          we_puts(" standard features -- an exceed claim is EARNABLE (add an axis the incumbent lacks).\n")
      }
  } else {
      we_puts("  VERDICT: NOT S-class -- BEHIND the incumbent by "); we_putn(outs[3])
      we_puts(" features. Those ABSENT features ARE the page builder's worklist (no overclaim).\n")
  }
  sys_exit(0); return 0
176}
nx_wiki_exceed.nx source

`nx_wiki_exceed.nx` source
