docs find-replace --format markdown leaks inline / block markdown as literal text (italic, bold, blockquote, fenced code)

[sebsnyk]

Summary

gog docs find-replace <docId> <find> <replace> --format markdown (and --content-file) successfully renders block-level structure but leaks inline / block formatting as literal markdown characters. The fix in #494 wired up the rendering path, but the converter itself doesn't handle several common markdown constructs.

What works

• ATX headings (#, ##, ###) → HEADING_1 / HEADING_2 / HEADING_3
• Bullet lists (- item)
• Pipe tables → TABLE element

What leaks as literal text

• Italic _text_ → underscores remain visible
• Bold **text** inside paragraphs → asterisks remain visible
• Blockquote > text → > lands as a literal character on its own line (breaks the callout)
• Fenced code blocks (```lang … ```) → flat plain paragraphs, no monospace, no grey background
• Inline emphasis combinations (e.g. **bold _and italic_**)

Repro (v0.16.0, macOS arm64)

    cat > /tmp/repro.md <<'EOF'
    ## Heading works
    
    This has **bold** and _italic_ and a `code span`.
    
    > blockquote should be a callout
    
    ```python
    print("hello")
    ```
    EOF
    
    DOC=$(gog docs create "gogcli md repro" --json | jq -r .id)
    gog docs write "$DOC" --text "REPLACE_ME"
    gog docs find-replace "$DOC" "REPLACE_ME" --content-file /tmp/repro.md --format markdown
    gog docs cat "$DOC"

Observed on a 30-page document today: 1 H1, 5 H2, 23 H3, 1 table all correct; every **…**, _…_, > …, and fenced code block leaked as raw markdown.

Expected

Match the fidelity of pasting the same .md into Google Docs UI via Edit > Paste from Markdown (which renders all of the above correctly).

Environment

• gog v0.16.0 (280eea7), macOS arm64
• Related: Docs write: Markdown not rendered when using API #272 (write path fix didn't cover find-replace converter), Bug: docs find-replace --format markdown silently does nothing #494 (find-replace path wired up but converter still incomplete on inline / block elements)

[steipete]

Fixed on main in 63ff5bd.

What changed:

• docs find-replace --format markdown now inserts stripped Google Docs text instead of leaking inline Markdown markers.
• Nested bold/italic, underscore emphasis, links, inline code, blockquotes, and fenced code blocks are covered by regression tests.
• Additional parser edge cases from review were fixed: code spans containing marker text, adjacent delimiter runs, whitespace-adjacent underscores, literal brackets before links, adjacent backticks, and literal trailing * after emphasis.

Proof:

• make ci passes locally.
• Live gog e2e with the test Google account created a temp doc, replaced REPLACE_ME from Markdown, read back rendered text without leaked markers, and trashed the temp doc.
• Codex review closeout returned no accepted/actionable findings.

Thanks again for the detailed repro.