Previously neither the module nor the name of the module of
pyfunctions were registered. This commit passes the module and
its name when creating a new pyfunction.
PyModule::add_function and PyModule::add_module have been added and are
set to replace `add_wrapped` in a future release. `add_wrapped` is kept
for compatibility reasons during the transition.
Depending on whether a `PyModule` or `Python` is the argument for the
Python function-wrapper, the module will be registered with the function.
Add benchmark where the sequential Rust version of the word count is
executed twice to demonstrate parallelism with Python threads. Also
slightly simplify the benchmark functions.
* Add `MODULE` const value to `PyTypeInfo` trait
* Allow setting module name in `pyclass` macro arguments
* Add `$module` argument to `pyobject_native_type` macro in `pyo3::types`
* Set the right module for builtin types in `pyo3::types`
* Add `module` argument to example `word_count.WordCounter` class
* Add `module` argument of `pyclass` macro to guide
* Reformat code using `rustfmt`
* Add tests and entry in CHANGELOG.md for `module` argument in `pyclass`
* Make `$module` parameter in `pyobject_native_type` optional
* Make `pyobject_native_type` declare module as `Some("builtins")` by default
Adding a build-backend key forces PEP 517 with pip >= 19.0, and so it
was left off while there was no PEP 517 support. Now that we have some
tentative PEP 517 support, we can explicitly add this key.
This drops "usedevelop", which does an editable installation of the
directory in favor of an isolated PEP 517/518 build environment. The
minimum version of tox has been bumped to one that has PEP 518 support.
This enables PEP 517 builds from an sdist, and should fix `tox` builds
using `pip >= 19.0`. This does not fix pip installing the directory;
pinning to `pip < 19.0` is still the best strategy for that.
* wip
* removed stuff
* removed another change
* implemented minimum amouth of ifdefs to make pypy3 hello world to compile
* implemented minimum amount of ifdefs to make pypy3 hello world to compile
* hacking on build.rs
* compiler is happy!
* few todos remain
* extracted build logic to seperate module
* added pypy test
* finally fixed pypy structs
* removed some todos
* test should now be machine independent
* fixed all pypy3 symbols
* added pypy feature
* removed `is_pypy`
* added pypy2 declerations also
* fix for cpython2
* improved libpypy detection
* added all pypy2 macros
* fixed errneous type
* more fixes
* fix python2 string macros
* modsupport symbol
* fix
* fixed and added many symbols
* fixes
* remove dup
* remove mac-specific config
* fix all name mangling macros
* unite imports
* missing symbol
* fix pybool
* implemented another missing symbol
* it works
* fix merge conflict
* uncomment non default features
* cargo.toml
* Cargo fmt
* small merge fixes
* use newer build version
* whoops
* fix build script
* more build hacks
* some random hiccups
* small fixes
* it builds!
* it builds and runs
* revert everything in FFI2
* revert changes to ffi2
* check python3 for pypy
* tiny fix
* revert ffi2 for real
* revert weird formatting changes
* bring back missing feature
* tiny error
* fix py3.7 issue
* add pypy3.5 6.0 to travis
* remove dbg!
* another tiny fix
* removed some useless annotations, and fixed inlines annotations
* removed `pretty_assertions`
* removed pypy feature from cargo.toml
* fix for Py_CompileStringFlags
* tox runs word_count!
* __dict__ changes are not supported for PyPy
* fix 3.7 and copy comment
* fix test script 😳
* transfer ownership of strings to cpython when possible
* remove cstr! macro
* added missing nuls
* as_bytes() -> b’’ string
* symbol removed by mistake
* properly shim pypy date time API, some tests are passing!
* extension_module tests now not crashing! (some still skipped)
* maybe travis has new pypy version?
* small error on windows (build script)
* fix conditional compilation
* try to make tests run on travis..
* invert condition
* added pytest-faulthandler to facilitate debugging
* correctly name dir
* use full paths
* say —yes to conda
* fix
* syntax error
* change PATH
* fixed a terrible bug with PyTypeObjects in PyPy
* fix PyTypeObject defs
* re-enabled tests!
* all tests are passing!
* make the fix ad-hoc for now
* removed build module
* revert changes that cause an additional GC bug
* prevented buggy test from failing pypy
* removed unused comment
* don’t run coverage on pypy
* removed some erroneous symbols from function calls which are actually macros
* restore py37 pyunicode missing def
* use only `link_name` in PyPy specific declarations
* only setup PyPy when testing against PyPy
* annotation that was eaten during merge
* remove change to comment by mistake + unnecessary changes to cargo.toml
* xfail dates test only on pypy
* changed comment to be a little more helpful
* cleaned up some warnings
* Update src/ffi3/ceval.rs
Co-Authored-By: omerbenamram <omerbenamram@gmail.com>
* @konstin PR notes
* rustfmt
* some documentation
* if configured via env var only, default to cpython
* remove extra unsafe
* refer users to guide for pypy
* Update guide/src/pypy.md
Co-Authored-By: omerbenamram <omerbenamram@gmail.com>
* Update guide/src/pypy.md
Co-Authored-By: omerbenamram <omerbenamram@gmail.com>
* @konstin applied patch
* check that pypy at least build
* search explicitly for libpypy
* added note about some known unsupported features
* use ld_version
* export PYTHON_SYS_EXECUTABLE to `cargo build` test
* inverted if
* always link pypy dynamically
* remove unused imports
* Apply @kngwyu’s suggestion
* fix tox configuration
* try conda virtualenv
* try to simply not install python at all inside pypy environment
* setup pypy before using “python"
* use system_site_packages
* revert change to .travis
* moved cpyext datetime documentation to module level, and revised it.
* Update src/ffi/datetime.rs
Co-Authored-By: omerbenamram <omerbenamram@gmail.com>
* rustfmt
* Update src/ffi/datetime.rs
Co-Authored-By: omerbenamram <omerbenamram@gmail.com>
* kept only notes that are relevant to users.
* invert if
* use bash and not sh
* Removed a lot of clutter, unified some code
* Started using syn::parse::Parse for pyfunction attributes
* No more newlines between imports
* Renamed `#[prop(get, set)]` to `#[pyo3(get, set)]`
* `#[pyfunction]` now supports the same arguments as `#[pyfn()]`
* Some macros now emit proper spanned errors instead of panics.
Because it's unlikely that anything other than the `year` parameter will
change in the C Python API, the rest can be restricted to their logical
ranges, which improves the compile-time error checking.
While the valid ranges for the constructor parameters is the same when
expressed as either u32 or i32, since the Python API uses i32 in their
public interface, we won't have to make any changes to the signatures if
the Python behavior changes (e.g. supporting negative years) without
their API changing.
Yuji Kanagawa