ludic/tools/ludic-cli/docgen_check.ludic

# docgen_check.ludic — `ludic dev docs-check [DIR]`: the coverage / integrity guard over
# a generated docs site, ported from tools/docgen/check.py. Fails (exit 1) if the
# pages contract is broken, an inventory symbol lacks a source file or page, a
# token is documented on two pages, a namespace/section is split across dirs, or
# a highlighter link target was not generated. Thin (un-expanded) symbols warn.

# a key -> set-of-values bucket (for token/dir conflict detection)
property Bucket { k1: pointer = "", k2: pointer = "", vals: []pointer }

function bucket_get(bs: []Bucket, k1: pointer, k2: pointer) -> Bucket {
  var i = 0
  while i < len(bs) {
    if bs[i].k1 == k1 and bs[i].k2 == k2 { return bs[i] }
    i += 1
  }
  let b = new Bucket
  b.k1 = k1
  b.k2 = k2
  b.vals = new []pointer
  push(bs, b)
  return b
}
function bucket_add(bs: []Bucket, k1: pointer, k2: pointer, v: pointer) -> void {
  let b = bucket_get(bs, k1, k2)
  var i = 0
  while i < len(b.vals) { if b.vals[i] == v { return };  i += 1 }
  push(b.vals, v)
}
# ", ".join(sorted(vals))
function join_sorted(vals: []pointer) -> pointer {
  strs_sort(vals)
  let b = sb_new()
  var i = 0
  while i < len(vals) {
    if i > 0 { sb_puts(b, ", ") }
    sb_puts(b, vals[i])
    i += 1
  }
  return sb_str(b)
}

# parse_front: like parse_doc but body is stripped on both ends (check.py)
function parse_front(path: pointer) -> Doc {
  let d = new Doc
  d.m = meta_new()
  let text = read_file(path)
  if text == null { d.body = "";  return d }
  d.body = text
  if not s_starts(text, "---") { return d }
  let end = s_index(text, "\n---", 3)
  if end < 0 { return d }
  let fm = sslice(text, 3, end)
  let fn2 = slen(fm)
  var j = 0
  while j < fn2 {
    let ln = line_at(fm, j)
    j = j + slen(ln) + 1
    let t = s_trim(ln)
    if slen(t) == 0 { continue }
    let c = s_index(ln, ":", 0)
    if c < 0 { continue }
    meta_put(d.m, s_trim(sslice(ln, 0, c)), s_trim(sslice(ln, c + 1, slen(ln))))
  }
  d.body = s_trim(sslice(text, end + 4, slen(text)))
  return d
}

# count Unicode codepoints (a byte is a lead unless (b & 0xC0) == 0x80)
function cp_len(s: pointer) -> int {
  let n = slen(s)
  var c = 0
  var i = 0
  while i < n { if (s[i] & 192) != 128 { c += 1 };  i += 1 }
  return c
}
function replace_all(hay: pointer, needle: pointer, repl: pointer) -> pointer {
  if slen(needle) == 0 { return hay }
  let b = sb_new()
  let nl = slen(needle)
  let n = slen(hay)
  var i = 0
  while i < n {
    if s_starts_at(hay, i, needle) { sb_puts(b, repl);  i += nl }
    else { sb_putc(b, hay[i]);  i += 1 }
  }
  return sb_str(b)
}
# strip a filename's trailing ".md" -> the fallback id (fn[:-3])
function drop_md(fn: pointer) -> pointer {
  let n = slen(fn)
  if n >= 3 { return sslice(fn, 0, n - 3) }
  return fn
}

function cmd_docs_check() -> int {
  var site = "build/pages"
  if arg_total() >= 3 { site = arg_n(2) }
  let problems = new []pointer
  let warnings = new []pointer

  # 1) pages contract
  if not file_exists(`{site}/index.html`) { push(problems, "missing required file: index.html") }
  if not file_exists(`{site}/api.html`) { push(problems, "missing required file: api.html") }
  if not file_exists(`{site}/.nojekyll`) { push(problems, "missing required file: .nojekyll") }
  # the landing page tells people to curl this; a site without it publishes a
  # broken install command
  if not file_exists(`{site}/install.sh`) { push(problems, "missing required file: install.sh") }

  # 2) coverage + thin detection over docs/language/**
  let id2src = meta_new()
  let thin = new []pointer
  let cats = list_sorted("docs/language")
  var ci = 0
  while ci < len(cats) {
    let cat = cats[ci]
    ci += 1
    let cdir = "docs/language/" + cat
    if not is_dir(cdir) { continue }
    let files = list_sorted(cdir)
    var fi = 0
    while fi < len(files) {
      let fn = files[fi]
      fi += 1
      if not s_ends(fn, ".md") { continue }
      if fn == "_section.md" { continue }
      let d = parse_front(cdir + "/" + fn)
      let sid = meta_get(d.m, "id", drop_md(fn))
      meta_put(id2src, sid, cdir + "/" + fn)
      # thin: drop a Parameters: block (to end, case-insensitive) + the tip line
      let tipline = meta_get(d.m, "tip", "")
      let low = lower_ascii(d.body)
      var btext = d.body
      let pp = s_index(low, "parameters:", 0)
      if pp >= 0 { btext = sslice(d.body, 0, pp) }
      btext = s_trim(btext)
      if slen(tipline) > 0 { btext = s_trim(replace_all(btext, tipline, "")) }
      if cp_len(btext) < 40 { push(thin, sid) }
    }
  }
  # coverage against the authoritative inventory
  let invtext = read_file("tools/docgen/inventory.json")
  if invtext != null {
    let inv = json_parse(invtext)
    if inv.t == JV_OBJ {
      var k = 0
      while k < len(inv.keys) {
        let cat = inv.keys[k]
        let ids = inv.kids[k]
        k += 1
        if ids.t != JV_ARR { continue }
        var m = 0
        while m < len(ids.kids) {
          let sid = ids.kids[m].s
          m += 1
          if meta_get(id2src, sid, "") == "" { push(problems, `no source file for inventory symbol: {sid} ({cat})`) }
          else if not file_exists(`{site}/{sid}.html`) { push(problems, `no generated page for symbol: {sid}.html`) }
        }
      }
    }
  }

  # 2b/2c) duplicate tokens + one-dir-per-namespace/section
  let tok2ids = new []Bucket    # k1=kind, k2=tok -> ids
  let ns2dirs = new []Bucket    # k1="", k2=ns -> dirs
  let id2dirs = new []Bucket    # k1="", k2=section id -> dirs
  let title2dirs = new []Bucket  # k1="", k2=lower(title) -> dirs
  ci = 0
  while ci < len(cats) {
    let cat = cats[ci]
    ci += 1
    let cdir = "docs/language/" + cat
    if not is_dir(cdir) { continue }
    let secp = cdir + "/_section.md"
    if file_exists(secp) {
      let sd = parse_front(secp)
      bucket_add(id2dirs, "", meta_get(sd.m, "id", cat), cat)
      bucket_add(title2dirs, "", lower_ascii(s_trim(meta_get(sd.m, "title", ""))), cat)
    }
    let files = list_sorted(cdir)
    var fi = 0
    while fi < len(files) {
      let fn = files[fi]
      fi += 1
      if not s_ends(fn, ".md") { continue }
      if fn == "_section.md" { continue }
      let d = parse_front(cdir + "/" + fn)
      let kind = meta_get(d.m, "kind", "")
      let id = meta_get(d.m, "id", "")
      let toks = split_ws(meta_get(d.m, "tokens", ""))
      var ti = 0
      while ti < len(toks) { bucket_add(tok2ids, kind, toks[ti], id);  ti += 1 }
      let ns = meta_get(d.m, "ns", "")
      if slen(ns) > 0 { bucket_add(ns2dirs, "", ns, cat) }
    }
  }
  var bi = 0
  while bi < len(tok2ids) {
    let b = tok2ids[bi]
    bi += 1
    if len(b.vals) > 1 { push(problems, `token '{b.k2}' ({b.k1}) documented on multiple pages: {join_sorted(b.vals)}`) }
  }
  bi = 0
  while bi < len(ns2dirs) {
    let b = ns2dirs[bi]
    bi += 1
    if len(b.vals) > 1 { push(problems, `namespace '{b.k2}' documented from multiple dirs: {join_sorted(b.vals)}`) }
  }
  bi = 0
  while bi < len(id2dirs) {
    let b = id2dirs[bi]
    bi += 1
    if len(b.vals) > 1 { push(problems, `section id '{b.k2}' declared by multiple dirs: {join_sorted(b.vals)}`) }
  }
  bi = 0
  while bi < len(title2dirs) {
    let b = title2dirs[bi]
    bi += 1
    if slen(b.k2) > 0 and len(b.vals) > 1 { push(problems, `section title '{b.k2}' shared by multiple dirs: {join_sorted(b.vals)}`) }
  }

  # 3) thin warning
  if len(thin) > 0 {
    strs_sort(thin)
    let b = sb_new()
    var i = 0
    while i < len(thin) and i < 12 {
      if i > 0 { sb_puts(b, ", ") }
      sb_puts(b, thin[i])
      i += 1
    }
    var tail = ""
    if len(thin) > 12 { tail = " …" }
    push(warnings, `{string(len(thin))} symbols still have only seed text: {sb_str(b)}{tail}`)
  }

  # 4) highlighter targets exist
  let sjtext = read_file(`{site}/symbols.json`)
  if sjtext != null {
    let sj = json_parse(sjtext)
    let hl = j_get(sj, "highlight")
    let targets = new []pointer
    collect_targets_str(j_get(hl, "keywords"), targets)
    collect_targets_str(j_get(hl, "types"), targets)
    collect_targets_str(j_get(hl, "phases"), targets)
    collect_targets_str(j_get(hl, "annotations"), targets)
    collect_targets_id(j_get(hl, "builtins"), targets)
    collect_targets_id(j_get(hl, "nsmethods"), targets)
    collect_targets_str(j_get(hl, "namespaces"), targets)
    strs_sort(targets)
    var i = 0
    while i < len(targets) {
      let tgt = targets[i]
      i += 1
      if not file_exists(`{site}/{tgt}.html`) { push(problems, `highlighter links to {tgt}.html but it was not generated`) }
    }
  }

  # report
  var wi = 0
  while wi < len(warnings) { print(`  warning: {warnings[wi]}`);  wi += 1 }
  if len(problems) > 0 {
    print("docs check FAILED:")
    var pi = 0
    while pi < len(problems) { print(`  - {problems[pi]}`);  pi += 1 }
    return 1
  }
  print(`docs check OK: {site}  ({string(len(id2src.keys))} symbols documented)`)
  return 0
}

# add every string value of an object's members (set semantics via the caller)
function collect_targets_str(o: JVal, out: []pointer) -> void {
  if o.t != JV_OBJ { return }
  var i = 0
  while i < len(o.kids) {
    let v = o.kids[i]
    i += 1
    if v.t == JV_STR { set_add(out, v.s) }
  }
}
# add each member value's ["id"] (builtins / nsmethods)
function collect_targets_id(o: JVal, out: []pointer) -> void {
  if o.t != JV_OBJ { return }
  var i = 0
  while i < len(o.kids) {
    let v = o.kids[i]
    i += 1
    let id = j_get(v, "id")
    if id.t == JV_STR { set_add(out, id.s) }
  }
}