Convenience functions related to modules and importing.
*Latest release 20220606*:
New import_extra(extra_package_name,distinfo) function to politely try to import a package which is associated with an extra.
## Function `direct_imports(src_filename, module_name=None)`
Crudely parse `src_filename` for `import` statements.
Return the set of directly imported module names.
If `module_name` is not `None`,
resolve relative imports against it.
Otherwise, relative import names are returned unresolved.
This is a very simple minded source parse.
## Function `import_extra(extra_package_name, distinfo)`
Try to import the package named `extra_package_name`
using `importlib.import_module` and return the imported package.
If an `ImportError` is raised,
riffle through the extras mapping in `distinfo['extras_requires']`
for the package name, and emit an informative warning
about the extras which require this package
and whose use a `pip install` time would bring the package in.
The original `ImportError` is then reraised.
If no extra is found this is presumed to be an error by the caller
and a `RuntimeError` is raised.
This function is for internal use as:
pkg = import_extra('some_package', DISTINFO)
which passes in the source module's `DISTINFO` mapping,
which I use as the basis for my package distributions.
A fuller example from my `cs.timeseries` module's
`plot` command line mode:
def cmd_plot(self, argv):
""" Usage: {cmd} datadir days fields...
"""
try:
import_extra('plotly', DISTINFO)
except ImportError as e:
raise GetoptError(
"the plotly package is not installed: %s" % (e,)
) from e
which produces this output:
timeseries.py: plot: import_extra('plotly'): package not available; the following extras pull it in: ['plotting']
timeseries.py: the plotly package is not installed: timeseries.py: plot: import_extra('plotly'): No module named 'plotly'
## Function `import_module_from_file(module_name, source_file, sys_path=None)`
Import a specific file as a module instance,
return the module instance.
Parameters:
* `module_name`: the name to assign to the module
* `source_file`: the source file to load
* `sys_path`: optional list of paths to set as `sys.path`
for the duration of this import;
the default is the current value of `sys.path`
Note that this is a "bare" import;
the module instance is not inserted into `sys.modules`.
*Warning*: `sys.path` is modified for the duration of this function,
which may affect multithreaded applications.
## Function `import_module_name(module_name, name, path=None, lock=None)`
Import `module_name` and return the value of `name` within it.
Parameters:
* `module_name`: the module name to import.
* `name`: the name within the module whose value is returned;
if `name` is `None`, return the module itself.
* `path`: an array of paths to use as sys.path during the import.
* `lock`: a lock to hold during the import (recommended).
## Function `module_attributes(M)`
Generator yielding the names and values of attributes from a module
which were defined in the module.
## Function `module_files(M)`
Generator yielding `.py` pathnames involved in a module.
## Function `module_names(M)`
Return a list of the names of attributes from a module which were
defined in the module.
# Release Log
*Release 20220606*:
New import_extra(extra_package_name,distinfo) function to politely try to import a package which is associated with an extra.
*Release 20210123*:
module_attributes: skip values from other modules _if we know the module_ (computed values like tuples have no module and still need to be returned).
*Release 20200521*:
* New import_module_from_file function to import a Python file as a module instance.
* New direct_imports(src_filename,module_name=None) returning the set of directly imports module names.
*Release 20190101*:
New functions: module_names, module_attributes.
*Release 20160918*:
* New generator function module_files yielding pathnames.
* import_module_name: accept name=None, just return the module.
* Add empty "install_requires" for DISTINFO completeness.
*Release 20150116*:
Initial PyPI release.