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

↩ module page · 321 lines · 20689 B
1// nx_papers_index.nx -- the Research Repository landing page, COMPILED FROM A REGISTRY (F1130).
2//
3// ★WHAT CHANGED AND WHY. This organ used to emit the index with one hardcoded paper_row()
4// call per paper, baked into the source. That is a rule-11 violation (data belongs in data,
5// not code): the list went stale on 2026-06-17, and adding a paper meant editing and
6// recompiling an organ. It also linked every paper into /wiki/, which is AUTH-GATED -- so
7// the "public research repository" was not publicly readable at all, and nothing that lives
8// only behind a login is citable.
9//
10// Now: rows come from a TSV registry and each row declares its own ACCESS. Adding a paper is
11// a data edit. Internal papers are listed honestly as access-controlled rather than rendered
12// as links that a reader cannot follow.
13//
14//   nx_papers_index <registry.tsv> <out.html>
15//   registry row: id \t access(public|internal) \t href \t title \t abstract \t result \t gates
16//
17// FAIL-CLOSED: an unreadable or empty registry REFUSES (exit 3) rather than emitting an empty
18// repository page -- a page saying "no papers" is indistinguishable from a broken read, and
19// silently publishing one would erase the repository.
20//
21// license_tier: ORIGINAL   No hw writes (Rule 26).   expect_exit: 0
22import "nx_syscalls.nx"
23import "nx_itoa_lib.nx"   // shared MSB-first emitter (zero-alloc)
24
25const PX_COLS: i64 = 7
26const PX_CAP: i64 = 262144
27const PX_OUT: i64 = 524288
28const PX_TAB: i64 = 9
29const PX_NL: i64 = 10
30
31func px_len(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } return n }
32func px_puts(s: *u8) -> i64 { sys_write(1, s, px_len(s)); return 0 }
33// MIGRATED to the shared emitter (debt 1785563586). The old body mmapped a scratch buffer
34// per call and never freed it. At PAGE granularity that is 4096B leaked PER CALL -- the
35// defect that took 28.5GB of a 36GB host in nx_ts_lumadiff (2MB input, ~3.66M calls).
36// nxi_* is MSB-first, allocates NOTHING, and emits identical bytes including the sign.
37func px_num(v: i64) -> i64 { nxi_out(v); return 0 }
38func px_read(path: *u8, buf: *u8, cap: i64) -> i64 {
  let fd: i64 = sys_openat_rd(path)
  if fd < 0 { return 0 - 1 }
  var n: i64 = 0
  var go: i64 = 1
  while go == 1 {
      let base: i64 = buf as i64
      let r: i64 = sys_read(fd, (base + n) as *u8, cap - n)
      if r <= 0 { go = 0 } else { n = n + r }
      if n >= cap { go = 0 }
  }
  sys_close(fd)
  return n
51}
52// append a NUL-terminated literal
53func px_app(dst: *u8, off: i64, s: *u8) -> i64 {
  var i: i64 = 0
  var o: i64 = off
  while s[i] != (0 as u8) { dst[o] = s[i]; o = o + 1; i = i + 1 }
  return o
58}
59// append src[a..b)
60func px_appn(dst: *u8, off: i64, src: *u8, a: i64, b: i64) -> i64 {
  var i: i64 = a
  var o: i64 = off
  while i < b { dst[o] = src[i]; o = o + 1; i = i + 1 }
  return o
65}
66func px_eol(buf: *u8, n: i64, i: i64) -> i64 {
  var e: i64 = i
  var s: i64 = 1
  while s == 1 {
      if e >= n { s = 0 } else { if buf[e] == (PX_NL as u8) { s = 0 } else { e = e + 1 } }
  }
  return e
73}
74// split line [ls,le) into PX_COLS ranges; col[c*2],col[c*2+1]. Returns columns found.
75func px_cols(buf: *u8, ls: i64, le: i64, col: *i64) -> i64 {
  var c: i64 = 0
  var p: i64 = ls
  while c < PX_COLS {
      var q: i64 = p
      var s: i64 = 1
      while s == 1 {
          if q >= le { s = 0 } else { if buf[q] == (PX_TAB as u8) { s = 0 } else { q = q + 1 } }
      }
      col[c*2] = p
      col[c*2+1] = q
      if q >= le {
          c = c + 1
          while c < PX_COLS { col[c*2] = le; col[c*2+1] = le; c = c + 1 }
          return c
      }
      p = q + 1
      c = c + 1
  }
  return c
95}
96func px_is(buf: *u8, a: i64, b: i64, lit: *u8) -> i64 {
  let l: i64 = px_len(lit)
  if b - a != l { return 0 }
  var i: i64 = 0
  while i < l { if buf[a+i] != lit[i] { return 0 } i = i + 1 }
  return 1
102}
103
104// Append a non-negative integer INTO the page buffer. px_num writes to stdout, which cannot serve a
105// masthead that states the repository's own counts -- and a hand-typed count in a generated index is
106// exactly the kind of number that goes stale silently.
107func px_appnum(out: *u8, pos: i64, v: i64) -> i64 {
  let t: *u8 = sys_mmap(28)
  var m: i64 = v
  var k: i64 = 0
  if m <= 0 { t[0] = 48 as u8; k = 1 } else {
      while m > 0 { t[k] = (48 + (m % 10)) as u8; m = m / 10; k = k + 1 }
  }
  var p: i64 = pos
  var i: i64 = 0
  while i < k { out[p] = t[k-1-i]; p = p + 1; i = i + 1 }
  return p
118}
119
120// Find `lit` inside buf[a..b); returns its start index, or -1.
121// WHY THIS EXISTS: almost every findings row in the registry contains a literal HONEST: clause -- the
122// author's own statement of where the work is behind. Rendered as one undifferentiated paragraph, that
123// clause reads as a footnote to a boast. Splitting on it gives the limitation its OWN channel, which is
124// the only page decision here that could not be lifted from a generic template: it is derived from the
125// estate's cardinal discipline that a caveat is a first-class result, not a disclaimer. A paper with no
126// such clause is then conspicuous by its absence, which is the pressure we want on the next author.
127func px_find(buf: *u8, a: i64, b: i64, lit: *u8) -> i64 {
  let l: i64 = px_len(lit)
  if l <= 0 { return 0 - 1 }
  if b - a < l { return 0 - 1 }
  var i: i64 = a
  while i <= b - l {
      var j: i64 = 0
      var m: i64 = 1
      while j < l { if buf[i+j] != lit[j] { m = 0; j = l } else { j = j + 1 } }
      if m == 1 { return i }
      i = i + 1
  }
  return 0 - 1
140}
141
142func main(argc: i64, argv: *i64) -> i64 {
  if argc < 3 {
      px_puts("usage: nx_papers_index <registry.tsv> <out.html>\n" as *u8)
      sys_exit(2); return 2
  }
  let reg: *u8 = sys_mmap(PX_CAP)
  let rn: i64 = px_read(argv[1] as *u8, reg, PX_CAP - 16)
  if rn <= 0 {
      // FAIL-CLOSED: never publish an empty repository over a real one.
      px_puts("PAPERS-INDEX REFUSED: registry unreadable or empty -> " as *u8)
      px_puts(argv[1] as *u8)
      px_puts("\n  (an empty index is indistinguishable from a broken read; refusing to emit)\n" as *u8)
      sys_exit(3); return 3
  }
  let out: *u8 = sys_mmap(PX_OUT)
  var o: i64 = 0
  o = px_app(out, o, "<!doctype html>\n<html lang=\"en\">\n<head>\n<meta charset=\"utf-8\">\n" as *u8)
  o = px_app(out, o, "<meta name=\"viewport\" content=\"width=device-width,initial-scale=1\">\n" as *u8)
  o = px_app(out, o, "<title>Nishi Research Repository</title>\n" as *u8)
  o = px_app(out, o, "<meta name=\"description\" content=\"Papers compiled from measured gates. Every quantitative claim cites a re-runnable gate.\">\n" as *u8)
  // ---- STYLESHEET (rebuilt 2026-08-06 to the estate's polish bar) --------------------------------
  // The palette is SEMANTIC, not decorative: proof/open/hold each carry a paired foreground+wash so a
  // paper's ACCESS STATE is legible as colour and not only as words. Three type roles do real work --
  // a serif for reading, a sans for chrome, and a mono reserved for identifiers, labels and figures,
  // which is what keeps a repository of measured claims from reading like a blog. Both themes are
  // authored, and the data-theme overrides win over the media query in BOTH directions.
  o = px_app(out, o, "<style>\n" as *u8)
  o = px_app(out, o, ":root{--ground:#EEEFEA;--sunk:#E4E6DF;--card:#F7F8F3;--ink:#191C1A;--ink-soft:#4B4F4A;--ink-faint:#767B73;--proof:#3D6350;--proof-wash:#E0EAE4;--open:#2B3F6E;--open-wash:#E1E5EF;--hold:#8A4436;--hold-wash:#F1E3DF;--rule:#D8D5C8;--edge:#B4AF9C;--serif:\"Iowan Old Style\",\"Palatino Linotype\",Palatino,\"Book Antiqua\",Georgia,serif;--sans:ui-sans-serif,system-ui,-apple-system,\"Segoe UI\",Roboto,sans-serif;--mono:ui-monospace,\"SF Mono\",\"Cascadia Mono\",\"Roboto Mono\",Consolas,monospace;--measure:68ch;--page:1080px}\n" as *u8)
  o = px_app(out, o, "@media (prefers-color-scheme:dark){:root{--ground:#13161A;--sunk:#0F1215;--card:#1B1E23;--ink:#E4E3DB;--ink-soft:#AEB0A7;--ink-faint:#7D807A;--proof:#84B096;--proof-wash:#19241D;--open:#94A8DC;--open-wash:#1D2534;--hold:#CE8577;--hold-wash:#2A1D1A;--rule:#33362F;--edge:#4A4D45}}\n" as *u8)
  o = px_app(out, o, ":root[data-theme=\"dark\"]{--ground:#13161A;--sunk:#0F1215;--card:#1B1E23;--ink:#E4E3DB;--ink-soft:#AEB0A7;--ink-faint:#7D807A;--proof:#84B096;--proof-wash:#19241D;--open:#94A8DC;--open-wash:#1D2534;--hold:#CE8577;--hold-wash:#2A1D1A;--rule:#33362F;--edge:#4A4D45}\n" as *u8)
  o = px_app(out, o, ":root[data-theme=\"light\"]{--ground:#EEEFEA;--sunk:#E4E6DF;--card:#F7F8F3;--ink:#191C1A;--ink-soft:#4B4F4A;--ink-faint:#767B73;--proof:#3D6350;--proof-wash:#E0EAE4;--open:#2B3F6E;--open-wash:#E1E5EF;--hold:#8A4436;--hold-wash:#F1E3DF;--rule:#D8D5C8;--edge:#B4AF9C}\n" as *u8)
  o = px_app(out, o, "*{box-sizing:border-box}body{margin:0;background:var(--ground);color:var(--ink);font-family:var(--sans);font-size:17px;line-height:1.65;-webkit-font-smoothing:antialiased}\n" as *u8)
  o = px_app(out, o, ".wrap{max-width:var(--page);margin:0 auto;padding:0 28px 110px}\n" as *u8)
  o = px_app(out, o, ".mast{padding:72px 0 36px;border-bottom:2px solid var(--ink)}\n" as *u8)
  o = px_app(out, o, ".mast-eyebrow{font-family:var(--mono);font-size:11px;letter-spacing:.16em;text-transform:uppercase;color:var(--ink-faint);margin:0 0 22px}\n" as *u8)
  o = px_app(out, o, "h1{font-family:var(--serif);font-weight:400;font-size:clamp(2.3rem,5.4vw,3.9rem);line-height:1.05;letter-spacing:-.015em;text-wrap:balance;margin:0 0 22px;max-width:18ch}h1 em{font-style:italic;color:var(--open)}\n" as *u8)
  o = px_app(out, o, ".thesis{font-family:var(--serif);font-size:clamp(1.08rem,2vw,1.3rem);line-height:1.5;color:var(--ink-soft);max-width:58ch;margin:0 0 30px}\n" as *u8)
  o = px_app(out, o, ".mast-meta{display:flex;flex-wrap:wrap;gap:10px 30px;font-family:var(--mono);font-size:11.5px;letter-spacing:.06em;text-transform:uppercase;color:var(--ink-faint)}.mast-meta b{color:var(--ink-soft);font-weight:600}\n" as *u8)
  o = px_app(out, o, ".paper{padding:34px 0;border-bottom:1px solid var(--rule);display:grid;grid-template-columns:92px 1fr;gap:0 20px}\n" as *u8)
  o = px_app(out, o, ".pid{font-family:var(--mono);font-size:11px;letter-spacing:.12em;color:var(--edge);padding-top:8px}.pbody{min-width:0}\n" as *u8)
  o = px_app(out, o, ".paper h2{font-family:var(--serif);font-weight:400;font-size:clamp(1.25rem,2.4vw,1.6rem);line-height:1.2;letter-spacing:-.01em;margin:0 0 10px;text-wrap:balance}\n" as *u8)
  o = px_app(out, o, ".paper h2 a{color:var(--ink);text-decoration-thickness:1px;text-underline-offset:3px}.paper h2 a:hover{color:var(--open)}\n" as *u8)
  o = px_app(out, o, ".tag{display:inline-block;font-family:var(--mono);font-size:9.5px;letter-spacing:.1em;text-transform:uppercase;font-weight:600;padding:3px 8px;border-radius:2px;margin-left:8px;vertical-align:.28em;white-space:nowrap;background:var(--hold-wash);color:var(--hold)}.tag.open{background:var(--open-wash);color:var(--open)}\n" as *u8)
  o = px_app(out, o, ".paper p{max-width:var(--measure);margin:0 0 12px;color:var(--ink-soft);font-size:.97rem}\n" as *u8)
  o = px_app(out, o, ".lbl{display:block;font-family:var(--mono);font-size:10px;letter-spacing:.14em;text-transform:uppercase;color:var(--ink-faint);margin-bottom:5px}\n" as *u8)
  o = px_app(out, o, ".res{border-left:3px solid var(--proof);padding-left:16px;margin:16px 0 14px;max-width:var(--measure)}.res p{margin:0;color:var(--ink)}\n" as *u8)
  o = px_app(out, o, ".limit{border-left:3px solid var(--hold);padding-left:16px;margin:14px 0;max-width:var(--measure)}.limit p{margin:0;color:var(--ink-soft);font-size:.93rem}.limit .lbl{color:var(--hold)}\n" as *u8)
  o = px_app(out, o, ".nolimit{font-family:var(--mono);font-size:10px;letter-spacing:.12em;text-transform:uppercase;color:var(--edge);margin:14px 0}\n" as *u8)
  o = px_app(out, o, ".gate{font-size:.86rem;color:var(--ink-faint)}.gate code{font-family:var(--mono);font-size:.92em}\n" as *u8)
  o = px_app(out, o, "a{color:var(--open);text-decoration-thickness:1px;text-underline-offset:2px}a:focus-visible{outline:2px solid var(--open);outline-offset:3px;border-radius:2px}\n" as *u8)
  o = px_app(out, o, "footer{margin-top:44px;padding-top:22px;border-top:2px solid var(--ink);color:var(--ink-faint);font-size:.86rem}footer p{max-width:var(--measure);margin:0 0 12px}footer strong{color:var(--ink-soft)}\n" as *u8)
  o = px_app(out, o, "@media (max-width:640px){.wrap{padding:0 20px 72px}.mast{padding-top:48px}.paper{grid-template-columns:1fr;gap:8px}.pid{padding-top:0}}\n" as *u8)
  // <main>, not <div>: nx_page_verify flagged the missing landmark when this was a bare div, and a
  // repository that grades everyone else's honesty does not get to ship an accessibility regression.
  o = px_app(out, o, "</style>\n</head>\n<body>\n<main class=\"wrap\">\n" as *u8)
  let col: *i64 = sys_mmap(PX_COLS * 2 * 8) as *i64
198
  // PRE-PASS. The masthead states the repository's OWN counts, so they must be known before it is
  // emitted. Counting the registry twice is cheaper than buffering the header, and it keeps the
  // figures DERIVED -- a hand-typed count inside a generated page is a number that goes stale in
  // silence, which is the one thing this repository exists to refuse.
  var pre_rows: i64 = 0
  var pre_pub: i64 = 0
  var pre_lim: i64 = 0
  var pi: i64 = 0
  while pi < rn {
      let ple: i64 = px_eol(reg, rn, pi)
      if ple > pi {
          px_cols(reg, pi, ple, col)
          if col[1] > col[0] {
              pre_rows = pre_rows + 1
              if px_is(reg, col[2], col[3], "public" as *u8) == 1 { pre_pub = pre_pub + 1 }
              if px_find(reg, col[10], col[11], "HONEST:" as *u8) >= 0 { pre_lim = pre_lim + 1 }
          }
      }
      pi = ple + 1
  }
219
  o = px_app(out, o, "<header class=\"mast\">\n<p class=\"mast-eyebrow\">Nishi Research &middot; compiled from measured gates</p>\n" as *u8)
  o = px_app(out, o, "<h1>Every number here <em>re-runs</em>.</h1>\n" as *u8)
  o = px_app(out, o, "<p class=\"thesis\">Each quantitative claim in this repository cites a re-runnable sovereign gate. No claim exceeds its measurement, every paper states plainly where it is behind, and this index is generated from a registry rather than hand-maintained.</p>\n" as *u8)
  o = px_app(out, o, "<div class=\"mast-meta\"><span><b>Papers</b> " as *u8)
  o = px_appnum(out, o, pre_rows)
  o = px_app(out, o, "</span><span><b>Publicly citable</b> " as *u8)
  o = px_appnum(out, o, pre_pub)
  o = px_app(out, o, "</span><span><b>Declaring a limitation</b> " as *u8)
  o = px_appnum(out, o, pre_lim)
  o = px_app(out, o, " of " as *u8)
  o = px_appnum(out, o, pre_rows)
  o = px_app(out, o, "</span><span><b>Compiled by</b> nx_papers_index</span><span><b>Source</b> registry, not hand-kept</span></div>\n</header>\n" as *u8)
  var i: i64 = 0
  var rows: i64 = 0
  var pub: i64 = 0
  while i < rn {
      let le: i64 = px_eol(reg, rn, i)
      if le > i {
          px_cols(reg, i, le, col)
          if col[1] > col[0] {
              rows = rows + 1
              let isPublic: i64 = px_is(reg, col[2], col[3], "public" as *u8)
              if isPublic == 1 { pub = pub + 1 }
              // The paper ID rides its own left rail rather than being glued to the title. That is a
              // structural device carrying real information -- a citable identifier -- not decoration,
              // and it lets the title be read as a sentence. BOTH access states get a semantic pill;
              // marking only the restricted ones leaves the reader inferring the default.
              o = px_app(out, o, "<section class=\"paper\">\n<div class=\"pid\">" as *u8)
              o = px_appn(out, o, reg, col[0], col[1])
              o = px_app(out, o, "</div>\n<div class=\"pbody\">\n<h2>" as *u8)
              if isPublic == 1 {
                  o = px_app(out, o, "<a href=\"" as *u8)
                  o = px_appn(out, o, reg, col[4], col[5])
                  o = px_app(out, o, "\">" as *u8)
                  o = px_appn(out, o, reg, col[6], col[7])
                  o = px_app(out, o, "</a><span class=\"tag open\">public</span>" as *u8)
              } else {
                  o = px_appn(out, o, reg, col[6], col[7])
                  o = px_app(out, o, "<span class=\"tag\">access-controlled</span>" as *u8)
              }
              o = px_app(out, o, "</h2>\n<p>" as *u8)
              o = px_appn(out, o, reg, col[8], col[9])
              o = px_app(out, o, "</p>\n" as *u8)
              // Split the findings at the author's own HONEST: marker. What precedes it is the result;
              // what follows is the limitation, and it gets its own block rather than trailing the
              // sentence that boasts. A row with no marker prints a quiet mono line saying so -- an
              // absent caveat should be VISIBLE, not indistinguishable from a modest one.
              let hpos: i64 = px_find(reg, col[10], col[11], "HONEST:" as *u8)
              var resEnd: i64 = col[11]
              var limStart: i64 = col[11]
              if hpos >= 0 {
                  // Trim the seam BOTH ways. Splitting on a marker leaves the result with a trailing
                  // space and the limitation with a leading one; rendered as separate blocks that reads
                  // as a typo rather than a cut. The marker is 7 bytes.
                  resEnd = hpos
                  var tw: i64 = 1
                  while tw == 1 {
                      if resEnd <= col[10] { tw = 0 } else {
                          if reg[resEnd-1] == (32 as u8) { resEnd = resEnd - 1 } else { tw = 0 }
                      }
                  }
                  limStart = hpos + 7
                  var lw: i64 = 1
                  while lw == 1 {
                      if limStart >= col[11] { lw = 0 } else {
                          if reg[limStart] == (32 as u8) { limStart = limStart + 1 } else { lw = 0 }
                      }
                  }
              }
              o = px_app(out, o, "<div class=\"res\"><span class=\"lbl\">Headline measured result</span><p>" as *u8)
              o = px_appn(out, o, reg, col[10], resEnd)
              o = px_app(out, o, "</p></div>\n" as *u8)
              if hpos >= 0 {
                  o = px_app(out, o, "<div class=\"limit\"><span class=\"lbl\">Where this is behind</span><p>" as *u8)
                  o = px_appn(out, o, reg, limStart, col[11])
                  o = px_app(out, o, "</p></div>\n" as *u8)
              } else {
                  o = px_app(out, o, "<p class=\"nolimit\">No limitation clause declared</p>\n" as *u8)
              }
              o = px_app(out, o, "<p class=\"gate\"><span class=\"lbl\">Reproducibility</span>" as *u8)
              o = px_appn(out, o, reg, col[12], col[13])
              o = px_app(out, o, " &mdash; sovereign gates (<code>nx_cc</code> to <code>nxasm</code>, no gcc); re-run to reproduce every number.</p>\n</div>\n</section>\n" as *u8)
          }
      }
      i = le + 1
  }
306
  o = px_app(out, o, "<footer>\n<p><strong>Doctrine.</strong> Compiled from measured gates; cite-or-refuse; no claim exceeds its gate; honest about where we are behind. For a non-novel capability the bar is stricter: the implementation is fully sovereign <em>and</em> the open step is a head-to-head against an independent reference, named rather than skipped.</p>\n" as *u8)
  o = px_app(out, o, "<p>Papers marked <em>access-controlled</em> live behind the internal wiki and are listed for completeness; they are not publicly citable until republished.</p>\n" as *u8)
  o = px_app(out, o, "<p>Generated by <code>nx_papers_index</code> from <code>knowledge/registry/papers_index.tsv</code>. Adding a paper is a data edit, not a code change.</p>\n</footer>\n</main>\n</body>\n</html>\n" as *u8)
310
  let fd: i64 = sys_openat_wr(argv[2] as *u8, 0x1a4)
  if fd < 0 { px_puts("PAPERS-INDEX IO: cannot open output\n" as *u8); sys_exit(4); return 4 }
  sys_write(fd, out, o)
  sys_close(fd)
  px_puts("PAPERS-INDEX OK rows=" as *u8); px_num(rows)
  px_puts(" public=" as *u8); px_num(pub)
  px_puts(" bytes=" as *u8); px_num(o)
  px_puts(" -> " as *u8); px_puts(argv[2] as *u8); px_puts("\n" as *u8)
  sys_exit(0)
  return 0
321}
nx_papers_index.nx source

`nx_papers_index.nx` source
