small perf improvements #1490
small perf improvements #1490
Conversation
sfc-gh-chu
requested review from
sfc-gh-dhuang,
sfc-gh-pmardziel and
sfc-gh-jreini
dosubot
bot
added
the
size:M
| @@ -378,7 +378,7 @@ def batch_insert_record( | |||
| self.orm.Record.parse(r, redact_keys=self.redact_keys) | |||
| for r in records | |||
| ] | |||
| session.bulk_save_objects(records_list) | |||
| session.add_all(records_list) | |||
There was a problem hiding this comment.
Whats difference between add_all and bulk_save objects?
There was a problem hiding this comment.
Functionally the same thing. bulk_save_objects is the legacy API in SQLAlchemy 2.0 so updated this to use the add_all API.
| for finfo in inspect.stack()[offset + 1 :]: | ||
| if not finfo.frame.f_globals["__name__"].startswith("trulens"): | ||
| return finfo.frame | ||
| gen = list(stack_generator(offset + 2)) |
There was a problem hiding this comment.
casting to a list shouldn't be necessary here right? and it's more memory efficient if you just iterate over the generator without forcing to evaluate the entire stack gen at once
There was a problem hiding this comment.
I was originally thinking the same. Then realized that because the generator is called lazily, its call stack would change and reflect the generator callee's call stack, not the call stack of external_caller_frame
There was a problem hiding this comment.
ah got it sg - could you add a small comment in the code?
There was a problem hiding this comment.
actually we may be able to get around this by passing in the current frame so that inspect.currentframe() is guarenteed to be invoked at the right time (pre-generator)
|
left a few minor comments. mostly LGTM |
* small perf improvements * nonetype check * write api * revert * stack_generator should take in callee's frame * skip first
Description
Other details good to know for developers
Please include any other details of this change useful for TruLens developers.
Type of change
not work as expected)