Skip to main content
Version: 2.25 (dev)

fmt

pants fmt [args]

Autoformat source code.

This goal runs tools that make 'syntactic' changes to source code, where the meaning of the code doesn't (usually) change.

See also:

  • The fix goal will run code-editing tools that may make semantic changes, not just syntactic ones.

  • The lint goal will validate code is formatted, by running these formatters and checking there's no change.

  • Documentation about formatters for various ecosystems, such as: Python, Go, JVM, Shell.

Backend: pants.core

Config section: [fmt]

Basic options​

only​

--fmt-only="['<str>', '<str>', ...]"
PANTS_FMT_ONLY
pants.toml
[fmt]
only = [
    '<str>',
    '<str>',
    ...,
]
default: []

Only run these formatters and skip all others.

The formatter names are outputted at the final summary of running this goal, e.g. isort and shfmt. You can also run fmt --only=fake to get a list of all activated formatters.

You can repeat this option, e.g. fmt --only=isort --only=shfmt or fmt --only=['isort', 'shfmt'].

Advanced options​

batch_size​

--fmt-batch-size=<int>
PANTS_FMT_BATCH_SIZE
pants.toml
[fmt]
batch_size = <int>
default: 128

The target number of files to be included in each formatter batch.

Formatter processes are batched for a few reasons:

  1. to avoid OS argument length limits (in processes which don't support argument files)
  2. to support more stable cache keys than would be possible if all files were operated on in a single batch.
  3. to allow for parallelism in formatter processes which don't have internal parallelism, or -- if they do support internal parallelism -- to improve scheduling behavior when multiple processes are competing for cores and so internal parallelism cannot be used perfectly.

In order to improve cache hit rates (see 2.), batches are created at stable boundaries, and so this value is only a "target" batch size (rather than an exact value).

Deprecated options​

None