Sorry, something went wrong.
|
BTW, @ncoghlan, do you think it's time to send this to the SC for review? We talked about it briefly in the "Faster CPython" team meeting today and we feel that this is a proposal we can get behind -- with our recommendation the SC might be slightly more inclined to accept it. (Although our recommendation is mostly based on thinking that everything this PEP proposes to change requires the user to have a frame object first, which basically means they are already in some kind of debugging/tracing mode, which we don't aim to optimize at all.) |
Sorry, something went wrong.
|
Aye, I actually recently wrote to Nathaniel to check his availability as PEP delegate. My own to do list for it at the moment is:
|
Sorry, something went wrong.
|
Okay, just be sure the draft implementation is based on the latest main branch. Things have changed a bit in this code... |
Sorry, something went wrong.
|
Yeah, the unwritten first task was "merge master to the PEP branch" |
Sorry, something went wrong.
This incorporates @markshannon's suggestion that the fast locals proxy only access the underlying frame storage, and not support writing additional keys to the dynamic snapshot stored in the C level f_locals struct field.
It also summarises the proposed CPython implementation changes, and cleans up quite a few old references to the idea of storing an instance of the fast locals proxy in the C level f_locals struct field (that hasn't been part of the proposal for ages, but it turned out there were still a lot of references to things working that way that I had missed).