extracting-tables

Extracting tables

Use this when the user wants tabular data out of HTML — pricing tables, data grids, spec sheets. html-to-markdown parses <table> elements into two surfaces at once: inline GFM Markdown tables in the content stream, and a structured tables array in the JSON output.

Two surfaces

# Inline GFM tables appear in the Markdown body
html-to-markdown input.html

# Structured table data appears under result.tables (JSON)
html-to-markdown --json input.html | jq '.tables'

• Markdown tables in content — | col | col | blocks, good for LLM ingestion and human reading.
• Structured tables array — each entry has a markdown field (pre-rendered) and a grid of structured cells (rows × cols). Use this when downstream code needs exact cell access.

Both are populated from the same parse; you do not need a flag to enable table parsing.

Extraction-only

When you only care about tables (not the Markdown body):

html-to-markdown --json --no-content input.html | jq '.tables'

Inspecting tables

# How many tables, and the row count of each
html-to-markdown --json input.html \
  | jq '.tables | to_entries | map({index: .key, rows: (.value.grid | length)})'

# Just the rendered markdown of the first table
html-to-markdown --json input.html | jq -r '.tables[0].markdown'

Line breaks in cells

By default <br> inside a cell is converted to a space. Keep hard breaks:

html-to-markdown input.html --br-in-tables

Programmatic access

from html_to_markdown import convert

result = convert(html)
for table in result.tables:
    print(table.markdown)      # rendered GFM markdown
    print(table.grid[0][0])    # first cell

import { convert } from "@kreuzberg/html-to-markdown";

// Node's convert() returns a JSON string — JSON.parse() it first.
const result = JSON.parse(convert(html));
for (const table of result.tables) {
  console.log(table.markdown);
}

Notes and limits

• Nested tables are flattened — the inner table is rendered inline within the parent cell.
• Merged / spanning cells are reconstructed positionally; the span itself is not preserved as separate metadata.
• Malformed tables still parse on a best-effort basis; check result.warnings (or --show-warnings) for non-fatal issues.

See ../html-to-markdown/references/cli-reference.md (Tables section) for the full flag set.