ueshin opened a new pull request, #47308:
URL: https://github.com/apache/spark/pull/47308
### What changes were proposed in this pull request?
Reduce the overhead of `_capture_call_site` by inline `inspect.stack` with
using `generator` instead of `list`.
Also, specify `context=0` for `inspect.getframeinfo` to avoid unnecessary
field retrievals.
### Why are the changes needed?
The `_capture_call_site` has inevitable overhead when `Column` operations
happen a lot, but we can reduce it.
E.g.,
```py
from functools import reduce
def alias_schema():
return df.select(reduce(lambda x, y: x.alias(f"col_a_{y}"), range(20),
F.col("a"))).schema
profile(alias_schema)
```
<img width="1106" alt="Screenshot 2024-07-11 at 15 24 31"
src="https://github.com/user-attachments/assets/1c677f56-86be-4e8f-9dd2-45c4c2c167f3";>
The function calls and duration are:
- before
```
18013240 function calls (18012760 primitive calls) in 2.327 seconds
ncalls tottime percall cumtime percall filename:lineno(function)
...
200 0.001 0.000 2.231 0.011
/.../python/pyspark/errors/utils.py:164(_capture_call_site)
```
- after
```
1421240 function calls (1420760 primitive calls) in 0.265 seconds
ncalls tottime percall cumtime percall filename:lineno(function)
...
200 0.001 0.000 0.182 0.001
/.../python/pyspark/errors/utils.py:165(_capture_call_site)
```
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
The existing tests.
### Was this patch authored or co-authored using generative AI tooling?
No.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
