Jpx Cli

Raw strings

jpx -r 'name' data.json                 # "alice" -> alice (no quotes)
jpx -j 'name' data.json                 # raw, no trailing newline

CSV / TSV

jpx --csv 'users[*]' data.json          # CSV with headers from object keys
jpx --tsv 'users[*]' data.json          # tab-separated
# Nested objects are flattened: {"addr":{"city":"NYC"}} -> addr.city column

Table

jpx -t 'users[*]' data.json             # formatted table
jpx -t --table-style ascii 'users[*]' data.json
# Styles: unicode (default), ascii, markdown, plain, rounded, sharp, modern

YAML / TOML

jpx -y 'config' data.json               # YAML output
jpx --toml 'config' data.json           # TOML output

Lines (NDJSON)

jpx -l 'users[*]' data.json             # one JSON value per line

Output to file

jpx -o result.json 'expression' data.json
jpx --csv -o data.csv 'users[*]' data.json

Input Modes

Slurp (-s)

Combine multiple JSON values from stdin into a single array:

echo -e '{"id":1}\n{"id":2}\n{"id":3}' | jpx -s 'length(@)'
# -> 3

Raw input (-R)

Read each line as a JSON string (not parsed as JSON):

echo -e "hello\nworld" | jpx -R -s '[*] | sort(@)'
# -> ["hello", "world"]

Null input (-n)

Use null as input, don't read stdin:

jpx -n 'now()'                    # current timestamp
jpx -n 'range(`1`, `11`)'         # [1,2,...,10]
jpx -n 'uuid()'                   # random UUID

Streaming Mode

Process NDJSON line by line with constant memory:

# Each line processed independently
cat events.ndjson | jpx --stream 'user.name'

# Null results are automatically skipped
cat events.ndjson | jpx --stream '[?level == `error`] | [0]'

# Streaming with CSV output (headers from first record)
cat events.ndjson | jpx --stream --csv '{user: user.name, action: event}'

# Streaming with TSV
cat events.ndjson | jpx --stream --tsv '@'

# Unbuffered output (flush after each record)
tail -f events.ndjson | jpx --stream --unbuffered 'message'

Note: --stream is incompatible with --table, --yaml, --toml, and --lines (these need the full dataset). Use --slurp instead if you need those formats.

See references/streaming.md for more streaming patterns.

Variable Binding

Bind variables for use in expressions:

# String variable
jpx --arg name Alice '[?name == $name]' data.json

# JSON variable (parsed)
jpx --argjson threshold 42 '[?score > $threshold]' data.json
jpx --argjson ids '[1,2,3]' '[?contains($ids, id)]' data.json

Query Files

Save and reuse expressions with .jpx query files:

# Run a named query
jpx -Q queries.jpx:active-users data.json

# Alternative syntax
jpx -Q queries.jpx --query active-users data.json

# List available queries
jpx -Q queries.jpx --list-queries

# Validate queries without running
jpx -Q queries.jpx --check

See references/query-files.md for the .jpx file format.

Function Discovery

jpx --list-functions                     # all 460+ functions
jpx --list-category String              # functions in a category
jpx --describe format_date              # detailed function info
jpx --search "date format"             # fuzzy search
jpx --similar group_by                  # find related functions

Data Inspection

jpx --stats -f data.json                # type, size, depth analysis
jpx --paths -f data.json                # all paths in dot notation
jpx --paths --types -f data.json        # paths with types
jpx --paths --values -f data.json       # paths with values

JSON Operations

# Diff two files (RFC 6902 JSON Patch)
jpx --diff before.json after.json

# Apply a patch
jpx --patch changes.json -f original.json

# Merge (RFC 7396)
jpx --merge overlay.json -f base.json

Other Features

# Explain expression (AST breakdown)
jpx --explain 'sort_by([*], &name) | [0]'

# Benchmark
jpx --bench 1000 'complex.expression' data.json

# Interactive REPL
jpx --repl -f data.json
jpx --repl --demo users                 # with demo dataset

# Exit status (for shell conditionals)
jpx -x '[?critical]' data.json && echo "found critical items"

Shell Pipeline Recipes

# Fetch API, extract, and format as table
curl -s https://api.example.com/users | jpx -t '[*].{name: name, email: email}'

# Process log file, count errors by type
cat app.log | jpx --stream -s '[*]' | jpx 'group_by(@, &error_type)'

# Convert JSON to CSV for spreadsheet
jpx --csv 'records[*]' -o export.csv data.json

# Find largest files in package.json dependencies
jpx 'dependencies | keys(@) | sort(@)' package.json

Related Skills

• jmespath-query -- JMESPath expression syntax
• jpx-functions -- 460+ extension functions
• jpx-mcp -- MCP server for AI agent integration