ability to execute precompiled sqlalchemy queries #294
Conversation
Codecov Report
@@ Coverage Diff @@
## master #294 +/- ##
==========================================
+ Coverage 92.79% 92.85% +0.05%
==========================================
Files 9 9
Lines 1110 1119 +9
Branches 158 161 +3
==========================================
+ Hits 1030 1039 +9
Misses 56 56
Partials 24 24
Continue to review full report at Codecov.
|
|
looks like i found a consistent way for sqlalchemy to implement this: will rework the PR and add method |
aiomysql/sa/connection.py
Outdated
| if not compiled: | ||
| compiled = query.compile(dialect=self._dialect) | ||
| if ( | ||
| dp and dp.keys() == compiled.params.keys() |
|
Sorry for late reply, after first look PR looks good, I will review more carefully this weekend. |
aiomysql/sa/connection.py
Outdated
| compiled = query.compile(dialect=self._dialect) | ||
| # parameters = compiled.params | ||
| if self._compiled_cache is not None: | ||
| key = (self._dialect, query) |
There was a problem hiding this comment.
I do not think we need self. _dialect as a part of the key, since it is always the same object.
There was a problem hiding this comment.
agree, a kind of overkill)
|
|
||
| def test_cache(self): | ||
| async def go(): | ||
| cache = dict() |
There was a problem hiding this comment.
Pure dict is not option for production system, since it has unbounded size. Should we also provide basic cache implementation also? something like LRU?
There was a problem hiding this comment.
I think this is up to implementor. LRU could be sufficient in most cases
BTW, sqlalchemy also does not provide any defaults - its only demand is that cache is some dict-like object
|
Thanks! |
|
new version available https://pypi.org/project/aiomysql/0.0.16/ |
Ability to execute pre-compiled queries removes ~30% sqlalchemy compilation overhead when the query is run repeatedly.