py/objmodule.c: Implement default weak links for all ports. #5229
Conversation
Previously ports chose whether or not to alias `ufoo` -> `foo`, and each port implemented this explicitly. This changes objmodule to always provide a `foo` weak link for any core `ufoo` module. This should be a no-op on any ports currently enabling weak links. `utime` and `uos` are provided by the ports, so they must still provide their own weak links for these modules. Fixes micropython#1740
da9b8ed to
e67c993
Compare
|
I strongly agree that the Unix port needs to be compatible to the other ports. It's a very useful tool for testing. |
e67c993 to
d744b43
Compare
|
Can you add a comment in the commit message about the effect this has (if any) on os and other libraries from micropython-lib installed via upip or similar? |
@stinos I am not aware of any meaningful effect this will have. I've updated the PR description. The only thing I can think of is if code uses the presence of (for example) |
|
Sorry it's just that I don't know how weak links are implemented so I was thinking of what happens for example when os from micropython-lib is installed. Will |
Weak links are effectively the same as having a set of symbolic links on the filesystem that is searched last. So an |
|
See also #4449 for related discussion. |
Agreed. The coverage build is better for testing/support because it includes all MicroPython features. I think the standard Unix build should be the same. |
|
See #5241 for an alternative to this. |
This commit implements automatic module weak links for all built-in
modules, by searching for "ufoo" in the built-in module list if "foo"
cannot be found. This means that all modules named "ufoo" are always
available as "foo". Also, a port can no longer add any other weak links,
which makes strict the definition of a weak link.
It saves some code size (about 100-200 bytes) on ports that previously had
lots of weak links.
Some changes from the previous behaviour:
- It doesn't intern the non-u module names (eg "foo" is not interned),
which saves code size, but will mean that "import foo" creates a new qstr
(namely "foo") in RAM (unless the importing module is frozen).
- help('modules') no longer lists non-u module names, only the u-variants;
this reduces duplication in the help listing.
Weak links are effectively the same as having a set of symbolic links on
the filesystem that is searched last. So an "import foo" will search
built-in modules first, then all paths in sys.path, then weak links last,
importing "ufoo" if it exists. Thus a file called "foo.py" somewhere in
sys.path will still have precedence over the weak link of "foo" to "ufoo".
See issues: #1740, #4449, #5229, #5241.
|
Superseded by #5241 |
This commit implements automatic module weak links for all built-in
modules, by searching for "ufoo" in the built-in module list if "foo"
cannot be found. This means that all modules named "ufoo" are always
available as "foo". Also, a port can no longer add any other weak links,
which makes strict the definition of a weak link.
It saves some code size (about 100-200 bytes) on ports that previously had
lots of weak links.
Some changes from the previous behaviour:
- It doesn't intern the non-u module names (eg "foo" is not interned),
which saves code size, but will mean that "import foo" creates a new qstr
(namely "foo") in RAM (unless the importing module is frozen).
- help('modules') no longer lists non-u module names, only the u-variants;
this reduces duplication in the help listing.
Weak links are effectively the same as having a set of symbolic links on
the filesystem that is searched last. So an "import foo" will search
built-in modules first, then all paths in sys.path, then weak links last,
importing "ufoo" if it exists. Thus a file called "foo.py" somewhere in
sys.path will still have precedence over the weak link of "foo" to "ufoo".
See issues: micropython#1740, micropython#4449, micropython#5229, micropython#5241.
This commit implements automatic module weak links for all built-in
modules, by searching for "ufoo" in the built-in module list if "foo"
cannot be found. This means that all modules named "ufoo" are always
available as "foo". Also, a port can no longer add any other weak links,
which makes strict the definition of a weak link.
It saves some code size (about 100-200 bytes) on ports that previously had
lots of weak links.
Some changes from the previous behaviour:
- It doesn't intern the non-u module names (eg "foo" is not interned),
which saves code size, but will mean that "import foo" creates a new qstr
(namely "foo") in RAM (unless the importing module is frozen).
- help('modules') no longer lists non-u module names, only the u-variants;
this reduces duplication in the help listing.
Weak links are effectively the same as having a set of symbolic links on
the filesystem that is searched last. So an "import foo" will search
built-in modules first, then all paths in sys.path, then weak links last,
importing "ufoo" if it exists. Thus a file called "foo.py" somewhere in
sys.path will still have precedence over the weak link of "foo" to "ufoo".
See issues: micropython#1740, micropython#4449, micropython#5229, micropython#5241.
add warmbit_bluepixel
I use the Unix port a lot for testing and keep tripping over this. I very strongly agree that the modules should be called
ufoobut also really like the weak linking behaviour so that code "just works" while still providing a nice mechanism for customisation and extension. After seeing a couple of recent forum posts about this, I looked through the history on #1740 and I'd like to propose that we just do this.This also avoids duplication in the ports and avoids a port "accidentally" missing a weak link (although I didn't notice any existing cases of this).
I'm not aware of any effect this will have on existing code. (Please correct me if I've missed anything here). If you're extending/overriding built-in modules (e.g. using
osfrom micropython-lib) then that will take precedence over the weak links, so will continue to function as it does currently.Previously ports chose whether or not to alias
ufoo->foo, and eachport implemented this explicitly. This changes objmodule to always provide a
fooweak link for any coreufoomodule.Other than now enabling weak links for the Unix port, this should be a
no-op on any ports currently enabling weak links.
utimeanduosare provided by the ports, so they must still providetheir own weak links for these modules.
Fixes #1740