typescore 0.10

#typescore

typescore generates typing completeness scores (and more) for a set of packages.

Usage:

  typescore [--packages <packages>] [--scores <scorefile>] [--sep <sep>] [--verbose] [<package>...]
  typescore --help
  typescore --version

Options:
  --packages <packages> File containing the list of packages.
  --scores <scorefile>  The output file (if not stdout).
  --sep <sep>           CSV column separator. [default: ,]
  -v, --verbose         Include package info in the output.
  -h, --help            Show this help.
  -V, --version         Show the version.

typescore uses pyright to score the typing completeness of a set of Python packages. It reads this list from <packages> and writes the results to <scorefile>. If errors prevent it from scoring a package it will set the score to 0%.

The output has the form:

package,typed,module,score,extra_columns

or, if --verbose is specified:

package,version,typed,module,score,stub_package,package_description,extra_columns

typed is a Boolean and tells whether the package had a py.typed file.

Note: we only score top-level modules, not submodules. The assumption is that scores for top-level modules would be reasonably representative of the packages all-up.

<packages> should have one package name per line. It can be a CSV file with the package name as the first column, in which case other columns will be included in the score file output (the extra_columns). A typical extra column might be the package rank on PyPI downloads.

While it would be useful to be able to measure the coverage scores on stub packages too, pyright does not support doing so. As a result, you should evaluate whether a stub package is better than the inline types for a package yourself beffore making use of it.