Load single builtin module #903
Conversation
that demonstrates this works.
|
|
||
| # First, load in many modules except quantum_mechanics. | ||
| _builtins = {} | ||
| import_and_load_builtins(exclude_files={"quantum_mechanics"}, clean_all=True) |
There was a problem hiding this comment.
Pytest loads all the tests before running them. So, as other unit tests load test.helper, and it calls import_and_load_builtins, when we reach this test, all the modules are already loaded. So, if we want to do a test over a "fresh" instance of the module, we need this second parameter.
|
In the back of my mind, I have the feeling that it does not make too much sense to store these lists and dictionaries outside of the Definitions object. In any case, I guess this PR is a good step forward. |
I agree that some of these should not exist and are signs of things that are still misfeatures in the code. Some information is used by the document system. The builtin operator precedence tables may be legitimate somehow, but not in this form. I don't mind having tables that exist that indicate various correspondences between SymPy and Mathics3. It might be somewhere else though. It has taken something like 3 or 4 iterations just to get to this stage, which is annoying and has been maddening and reflects how ill-designed things have been, make worse by "incremental" changes. It reinforces why going forward we should be taking a more high-level view inspired more by existing patterns for interpreters and Python coding. |
Inch forward towards being able to reduce startup time by loading only essential module.
The change here is that we can now import a single module by import name, e.g.
mathics.builtin.quantum_mechanics.Note that frontends may need to reload sessions, which is not desirable.