Re-apply type annotations from .pyi stubs to your codebase.
retype [OPTIONS] [SRC]... Options: -p, --pyi-dir DIRECTORY Where to find .pyi stubs. [default: types] -t, --target-dir DIRECTORY Where to write annotated sources. [default: typed-src] -i, --incremental Allow for missing type annotations in both stubs and the source. -q, --quiet Don't emit warnings, just errors. -a, --replace-any Allow replacing Any annotations. --hg Post-process source files to preserve implicit byte literals. --traceback Show a Python traceback on error --version Show the version and exit. --help Show this message and exit.
When you run
retype, it goes through all files you passed as SRC,
finds the corresponding .pyi files in the
types/ directory, and
re-applies typing annotations from .pyi to the sources, using the
Python 3 function and variable annotation syntax. The resulting
combined sources are saved in
You can also pass directories as sources, in which case
look for .py files in them recursively.
It's smart enough to do the following:
- reapply typing imports
- reapply function argument annotations
- reapply function return value annotations
- reapply method argument and return value annotations
- reapply function-level variable annotations
- reapply module-level name annotations
- reapply module-level type aliases
- reapply class-level field annotations
- reapply instance-level field annotations
- validate existing source annotations against the .pyi file
- validate source function signatures against the .pyi file
- read function signature type comments in .pyi files
- read variable type comments in .pyi files
- consider existing source type comments as annotations
- remove duplicate type comments from source when annotations are applied
- normalize remaining type comments in the source to annotations; this is done even if the corresponding .pyi file is missing
List of things to be done
- add a --backward option to output type comments instead of annotations
- handle if sys.version_info and sys.platform checks in stubs
- it's okay for a given .pyi file to be incomplete (gradual typing, baby!)
- it's okay for functions and classes to be out of order in .pyi files and the source
- it's an error for a function or class to be missing in the source
- it's an error for a function's signature to be incompatible between the .pyi file and the source
- it's an error for an annotation in the source to be incompatible with the .pyi file
- Line numbers in the annotated source will no longer match original source code; this is because re-application of types requires copying typing imports and alias definitions from the .pyi file.
- While formatting of the original source will be preserved, formatting of the applied annotations might differ from the formatting in .pyi files.
- The source where type annotations get re-applied cannot use the
- Class attribute annotations in
__init__()methods are moved verbatim to the respective
__init__()method in the implementation. They are never translated into class-level attribute annotations, so if that method is missing, the translation will fail. Similarly, class-level attribute annotations are never applied to
- Forward references in .pyi files will only be properly resolved for
type aliases and type vars (by inserting them right before they're
used in the source). Other forms of forward references will not work
in the source code due to out-of-order class and function definitions.
Modify your .pyi files to use strings.
retypewill not automatically discover failing forward references and stringify them.
- Local variable annotations present in the .pyi file are transferred to
the body level of the given function in the source. In other words,
if the source defines a variable within a loop or a conditional
retypewill create an value-less variable annotation at the beginning of the function. Use a broad type and constrain types in relevant code paths using
- Because of the above, existing source variable annotations and type
comments buried in conditionals and loops will not be deduplicated
mypywill complain that a name was already defined).
- An async function in the stub will match a regular function of the
same name in the same scope and vice versa. This is to enable
annotating async functions spelled with
python setup.py test
OMG, this is Python 3 only!
Relax, you can run retype as a tool perfectly fine under Python 3.6+ even if you want to analyze Python 2 code. This way you'll be able to parse all of the new syntax supported on Python 3 but also effectively all the Python 2 syntax at the same time.
By making the code exclusively Python 3.6+, I'm able to focus on the quality of the checks and re-use all the nice features of the new releases (check out pathlib or f-strings) instead of wasting cycles on Unicode compatibility, etc.
Note: to retype modules using f-strings you need to run on Python 3.6.2+ due to bpo-23894.
support --replace-any to allow replacing pre-existing
Anyannotations without raising errors
bugfix: don't re-apply
# type: ignoreas an annotation if followed by another comment. Original patch by Shannon Zhu.
bugfix: don't try to re-apply
# type: ignoreas a function annotation
bugfix: support arbitrary source file encodings, patch by Michael Overmeyer.
bugfix: support missing newlines at the end of the file, patch by Michael Overmeyer.
bugfix: in --incremental, format default values according to PEP 8 (no spaces around the = sign if the type is missing)
- bugfix: --incremental didn't work with multiple arguments before
- support --incremental stub application (i.e. allow for both stubs and the source to be missing annotations for some arguments and/or return value)
support async functions
support --traceback for getting more information about internal errors
first published version