regen.mde 0.8.0 → 0.8.2

package/scripts/run-smoke.ps1

@@ -1,28 +1,28 @@
-param(
-  [Parameter(Mandatory=$true)]
-  [string]$Docx,
-
-  [string]$Out = ".tmp\smoke",
-
-  [ValidateSet("assets", "base64", "s3")]
-  [string]$Images = "assets",
-
-  [string]$Config = ""
-)
-
-$ErrorActionPreference = "Stop"
-$root = Split-Path -Parent $PSScriptRoot
-$buildCorpus = Join-Path $root "bin\build-corpus.js"
-$katex = Join-Path $root "bin\validate-katex.js"
-$assetValidator = Join-Path $root "src\build_corpus\validate_assets.py"
-
-$args = @($buildCorpus, $Docx, "--out", $Out, "--images", $Images)
-if ($Config) {
-  $args += @("--config", $Config)
-}
-
-node @args
-node $katex $Out | Set-Content -Path (Join-Path $Out "katex-report.json") -Encoding UTF8
-py -3 $assetValidator $Out | Set-Content -Path (Join-Path $Out "asset-report.json") -Encoding UTF8
-
-Write-Host "Smoke test passed: $Out"
+param(
+  [Parameter(Mandatory=$true)]
+  [string]$Docx,
+
+  [string]$Out = ".tmp\smoke",
+
+  [ValidateSet("assets", "base64", "s3")]
+  [string]$Images = "assets",
+
+  [string]$Config = ""
+)
+
+$ErrorActionPreference = "Stop"
+$root = Split-Path -Parent $PSScriptRoot
+$buildCorpus = Join-Path $root "bin\build-corpus.js"
+$katex = Join-Path $root "bin\validate-katex.js"
+$assetValidator = Join-Path $root "src\build_corpus\validate_assets.py"
+
+$args = @($buildCorpus, $Docx, "--out", $Out, "--images", $Images)
+if ($Config) {
+  $args += @("--config", $Config)
+}
+
+node @args
+node $katex $Out | Set-Content -Path (Join-Path $Out "katex-report.json") -Encoding UTF8
+py -3 $assetValidator $Out | Set-Content -Path (Join-Path $Out "asset-report.json") -Encoding UTF8
+
+Write-Host "Smoke test passed: $Out"

package/src/build_corpus/__init__.py

@@ -1,3 +1,3 @@
-"""DOCX to Markdown conversion with native image and equation handling."""
-
-__version__ = "0.8.0"
+"""DOCX to Markdown conversion with native image and equation handling."""
+
+__version__ = "0.8.0"

package/src/build_corpus/docx_exporter.py

@@ -36,7 +36,7 @@ except ImportError:  # pragma: no cover - script-style invocation
 # like "\$252.3B" — are NOT mistaken for inline-math fences. Escaped dollars then
 # flow through the plain-text path and are unescaped to "$" (counted as a fixup).
 INLINE_TOKEN_RE = re.compile(
-    r"(!\[[^\]]*\]\([^)]+\)|\[[^\]]+\]\([^)]+\)|`[^`]+`|(?<!\\)\$\$[^$]+\$\$|(?<!\\)\$[^$\n]+\$|\*\*\*.+?\*\*\*|\*\*.+?\*\*|\*.+?\*)"
+    r"(!\[[^\]]*\]\([^)]+\)|\[[^\]]+\]\([^)]+\)|`[^`]+`|(?<!\\)\$\$(?:\\\$|[^$])+\$\$|(?<!\\)\$(?:\\\$|[^$\n])+\$|\*\*\*.+?\*\*\*|\*\*.+?\*\*|\*.+?\*)"
 )
 
 # Image targets python-docx can never embed as a picture (need an external
@@ -83,10 +83,16 @@ def count_input_elements(markdown: str) -> dict:
         line_text = lines[index]
         # inline elements on this line
         counts["images"] += len(re.findall(r"!\[[^\]]*\]\([^)]+\)", line_text))
+        # Math inside a link/image/code token is rendered as part of that token —
+        # INLINE_TOKEN_RE matches links/images/code BEFORE inline math — so strip
+        # those spans before counting equations to mirror the renderer. (e.g. a TOC
+        # entry "[... Allocation ($\alpha$) 117](#anchor)" is one link, not an equation.)
+        eq_text = re.sub(r"!\[[^\]]*\]\([^)]+\)|\[[^\]]+\]\([^)]+\)|`[^`]+`", " ", line_text)
         # display $$...$$ embedded in a line with surrounding text (render tokenizes these too)
-        counts["equations"] += len(re.findall(r"(?<!\\)\$\$[^$]+\$\$", line_text))
-        # inline math: standalone $...$ not part of $$ and not escaped
-        counts["equations"] += len(re.findall(r"(?<!\\)(?<!\$)\$[^$\n]+\$(?!\$)", line_text))
+        # (?:\\\$|[^$]) lets an escaped \$ — e.g. currency inside math — stay inside the fence
+        counts["equations"] += len(re.findall(r"(?<!\\)\$\$(?:\\\$|[^$])+\$\$", eq_text))
+        # inline math: standalone $...$ not part of $$ and not escaped (escaped \$ stays inside)
+        counts["equations"] += len(re.findall(r"(?<!\\)(?<!\$)\$(?:\\\$|[^$\n])+\$(?!\$)", eq_text))
         # links that are not images
         counts["links"] += len(re.findall(r"(?<!!)\[[^\]]+\]\([^)]+\)", line_text))
         index += 1