Processing Markdown

Processing Markdown with mq

Core Selectors

Selector Calls (Filtered Matching)

.h(1)          # Only h1 headings
.h(2, 3)       # h2 and h3 headings
.h(1..3)       # h1 through h3 (range)
.code("rust")  # Only Rust code blocks

Key Attribute Access

.h.level / .h.depth   # Heading level (1–6)
.h.value              # Heading text
.code.lang            # Code language
.code.value           # Code content
.link.url             # Link URL
.image.alt            # Image alt text
.list.checked         # Checkbox state (boolean)
."key"                # Dict key access (property selector)

Update Operator

.code.lang |= "rust"           # Change code language in-place
.link.url  |= "https://new"    # Update link URL

Recursive Descent

..ident    # Recursively select matching nodes in nested structures

Common Patterns

# Extract
mq '.h' file.md                              # All headings
mq '.h(2)' file.md                           # h2 only
mq '.code("rust")' file.md                  # Rust code blocks
mq '.link.url' file.md                       # All URLs
mq '.yaml | to_text' post.md              # Frontmatter

# Filter
mq 'select(.code)' file.md                  # Only code blocks
mq 'select(!.code)' file.md                 # Exclude code blocks
mq 'select(.h.level <= 2)' file.md          # h1 and h2 only
mq 'select(contains("TODO"))' file.md       # Nodes with "TODO"

# Transform
mq '.h | to_text' file.md                 # Headings as plain text
mq -U '.code.lang |= "rust"' file.md        # Update in place

# Multi-file
mq -A 'pluck(.code.value)' *.md             # Collect code values, per file
mq --eval-all -A '.h | to_text' *.md        # Combine all files into one query (e.g. cross-file TOC)
mq -S 's"\n---\n"' 'identity' *.md       # Merge with separator

# mq accepts multiple file args directly (shell glob expansion) —
# no need to loop over files in bash:
mq '.h | to_text' *.md work/*.md docs/*.md

# Format conversion
mq -F html 'identity' file.md             # Markdown → HTML
mq -F json '.h | to_text' file.md         # Headings → JSON
mq -I html 'identity' page.html           # HTML → Markdown

# Streaming large files
mq --stream 'select(contains("ERROR"))' large.md

# Section operations (needs -A or `nodes` — see EXAMPLES.md)
mq -A 'section::section("Installation")' file.md                                                             # Extract section by heading
mq -A 'section::filter_sections(fn(s): section::title(s) != "Deprecated";) | section::collect()' file.md    # Delete section by heading

HTML Input: Always Use Markdown Selectors

When using -I html, mq converts HTML to Markdown first — use Markdown selectors, not HTML tags.

# WRONG
curl -s https://example.com | mq -I html '.p | to_text'

# CORRECT
curl -s https://example.com | mq -I html '.text | to_text'
curl -s https://example.com | mq -I html '.link.url'
curl -s https://example.com | mq -I html '.h | to_text'

Essential CLI Flags

A small, stable cheat sheet — not exhaustive. See below for everything else.

For the full CLI option list (all flags, possible format values, auto-parsing by file extension, ARGS handling), run mq --help. For docs on a specific function or selector (signature, types, examples, required capability), run mq help <name> (e.g. mq help map, mq help .h1); pass --json for machine-readable output. Run mq help with no name to list every known function/selector.

Note: --args also accepts the hidden aliases --arg and --define (not shown in mq --help).

Node Attribute Reference

These attributes are Markdown-selector-specific and are not covered by mq help / mq --help.

Function Call Syntax

• All function calls require parentheses ().
• If a function is called with missing arguments, the piped value (|) is used as the first argument.

Environment Variables

• __FILE__ — full path to the file being processed
• __FILE_NAME__ — filename without path
• __FILE_STEM__ — filename without extension

For advanced examples, see EXAMPLES.md.

When NOT to Use mq

• Binary file processing
• Simple cat / echo with no transformation
• Non-Markdown data where jq (JSON) or yq (YAML) fits better