MARK IT DOWN
Mark It Down
页面 · Mark It Down

Sortable Tables + Per-Table Export

Status: shipped in Phase 0.5 · Issue: #5

Markdown tables in View mode are now sortable (click any column header) and ship a per-table export toolbar with CSV, TSV, and Excel (.xlsx) outputs. No new framework dep — sort + serialization are hand-rolled; xlsx uses SheetJS for the binary workbook format.

At a glance

WhereAny markdown table in View mode
SortClick a column header → asc → desc → none (original order)
Numeric vs stringBoth columns are detected per-cell; mixed columns fall back to localeCompare with numeric-aware sensitivity
Export formatsCSV, TSV, Excel (.xlsx) — buttons in a toolbar above each table
Savevscode.window.showSaveDialog defaulting to the markdown file's directory; info toast with Open / Reveal follow-ups

What ships

Sortable headers

Every <th> gets a click handler that cycles its data-sort attribute through none → asc → desc → none. The first sort stamps data-mid-original on each row so the "none" state can restore the original markdown order.

Comparison strategy:

text

This handles "$1,200" vs "$300" correctly, sorts "2025-04-29" alphabetically (which is also chronologically correct for ISO dates), and falls back gracefully on free-form text.

Indicator glyphs in the header:

  • ⇅ — unsorted (default)
  • ▲ — ascending
  • ▼ — descending

Only one column can be active at a time — clicking a new header resets the previous indicator.

Per-table toolbar

Above each table, a mini-toolbar:

Table 3 [CSV] [TSV] [Excel]

The label numbers each table on the page (1-indexed) for use as the default save filename. Buttons are scoped per table — clicking CSV on Table 3 only exports Table 3's rows.

CSV / TSV serialization

Hand-rolled in the webview:

  • Cells containing the separator, double-quote, or newline are quoted; embedded " is doubled per RFC 4180
  • Header row is included
  • Line terminator is \n (Excel + LibreOffice + sheets all accept this)
  • UTF-8 encoded by the host on write

Excel (.xlsx)

Uses SheetJS (xlsx package) — the standard JS spreadsheet library. The xlsx format is a zip of XML files; hand-rolling it would be a substantial undertaking and a maintenance liability. SheetJS adds ~600KB to the webview bundle (8.1MB → 8.7MB) which is the trade-off. If bundle size becomes a concern in v1+, we can lazy-load the xlsx module on first export.

Sort state and export interaction

Exports use the current row order in the DOM. So if you sort by "Cost (desc)" then export to CSV, the CSV reflects that order. To export the original markdown order, click the column header until it cycles back to none (or simply don't sort first).

Edge cases & limitations

  • Tables without <tbody>: marked emits <tbody> for every markdown table, so this is rare. The sort handler defends with if (!tbody) return; and exports skip such tables silently.
  • Cells with embedded HTML: cell.textContent strips tags for both sort and export. Bold/italic content sorts and exports as the unstyled text.
  • Multiple <thead> rows: only the first row is wired for sorting. Multi-header tables are uncommon in markdown anyway.
  • Tables inside blockquotes / lists: detected by main.viewing table selector, so the toolbar attaches consistently regardless of nesting.
  • Mermaid diagrams that contain tables: the attachTableActions() runs after renderView() and matches DOM tables under main.viewing. Mermaid SVG tables are inside <svg> and not matched — so they're not sortable, which is the intended behavior.
  • Very large tables (>1k rows): sort + export still work but a single click may block the main thread for a noticeable beat. Out of scope for v0.5; tracked as a future-work seed for chunked sort if reports come in.
  • Per-cell formulas (xlsx): cells are written as plain values via XLSX.utils.aoa_to_sheet. No formula reconstruction — markdown can't express formulas anyway.

Files of interest