gh-95913: Edit, expand & format Bytecode sect in 3.11 WhatsNew #98559
Conversation
CAM-Gerlach
added
3.11
CAM-Gerlach
requested review from
pablogsal,
brandtbucher,
ezio-melotti and
JelleZijlstra
Doc/whatsnew/3.11.rst
Outdated
| Changed Opcodes | ||
| --------------- | ||
|
|
||
| * Changed :opcode:`JUMP_IF_TRUE_OR_POP` and :opcode:`JUMP_IF_FALSE_OR_POP` | ||
| to now be relative rather than absolute. |
There was a problem hiding this comment.
I left the Changes section as it was for now (aside from minor technical edits), but there were a handful of additional Opcode changes of what seemed to be roughly comparable magnitude to those mentioned that weren't noted here, mostly grouped by a few specific common alterations, that we could include in a final short table (with columns Opcode(s) and Change), if you feel its worth including.
Here are the additional changes that could be mentioned:
List of additional changes
Added where oparg
- GET_AWAITABLE
Exception representation on the stack now consist of one, not three, items
- END_ASYNC_FOR
- POP_EXCEPT
- RERAISE
- WITH_EXCEPT_START
The exit function is in position 4 of the stack rather than 7.
- WITH_EXCEPT_START
If the low bit of namei is set, then a NULL is pushed to the stack before the global variable.
- LOAD_GLOBAL
i is no longer offset by the length of co_varnames
- LOAD_CLOSURE
- LOAD_DEREF
- LOAD_CLASSDEREF
- STORE_DEREF
- DELETE_DEREF
|
This should be also checked with @markshannon, @brandtbucher, @iritkatriel and @ericsnowcurrently. If this doesn't land today, it won't be included in 3.11.0 |
CAM-Gerlach
requested review from
iritkatriel,
markshannon and
ericsnowcurrently
and removed request for
JelleZijlstra
|
|
||
| * :opcode:`JUMP_IF_TRUE_OR_POP` and :opcode:`JUMP_IF_FALSE_OR_POP` are now | ||
| relative rather than absolute. | ||
| .. _whatsnew311-added-opcodes: |
There was a problem hiding this comment.
The list we have now has all the jump related changes together, but if you split it into a "added" table and a "replaced" table then they are no longer together. I don't know if that's an improvement.
There was a problem hiding this comment.
I created an alternative PR -#98566. It tries to consolidate related items into a high level descriptions, rather than enumerate all the changes. Maybe this is more useful for whatsnew. The full details, if you need them, as in the dis doc.
I added an item about the exc-info related changes. Didn't do anything about these:
GET_AWAITABLE
WITH_EXCEPT_START
If the low bit of namei is set, then a NULL is pushed to the stack before the global variable.
LOAD_GLOBAL
i is no longer offset by the length of co_varnames
LOAD_CLOSURE
LOAD_DEREF
LOAD_CLASSDEREF
STORE_DEREF
DELETE_DEREF
There was a problem hiding this comment.
Great points, thanks. I've incorporated the changes suggested via your PR, specifically:
- Consolidated the jump opcode additions, replacements and changes (except for the
NO_INTERRUPTopcode addition, which didn't appear to be closely related to the others) into a single Replaced table entry with a linked note - Just linked the high-level jump opcode name patterns, instead of every opcode.
- Added a "Changed" note about the exception stack representation
- (
EXC/EGchanges were already consolidated)
Beyond that, I've given this PR a major revamp based on your feedback above:
- Converted the "Added" section back to (much shorter) bullets
- Inlined the previous separate notes into the Replaced table
- Combined the previous "Removed" and "Changed" sections into one, condensing the Removed items into a single bullet to take up much less space
- Made additional textual refinements, particularly in the Changed entries, to be both more clear and more concise
In total, this retains nearly the same space consumption of the previous dense wall of flat bullets format (once allowing for the missing substantive additions, replacements and removals added here), while being much easier for the reader to quickly navigate, find what they're looking for and jump to more information in the dis docs.
| Flat Bullets | Initial PR | Revised PR |
|---|---|---|
 |
 |
 |
|
Yeah this is better! Just CHECK_EG_MATCH is also new for PEP 654. |
|
Heads up: please land this before 3Pm GMT+0 if you want it to be in 3.11.0 |
|
Thanks @CAM-Gerlach for the PR, and @iritkatriel for merging it 🌮🎉.. I'm working now to backport this PR to: 3.11. |
…ythonGH-98559) (cherry picked from commit 22739a0) Co-authored-by: C.A.M. Gerlach <CAM.Gerlach@Gerlach.CAM>
|
GH-98597 is a backport of this pull request to the 3.11 branch. |
) (cherry picked from commit 22739a0) Co-authored-by: C.A.M. Gerlach <CAM.Gerlach@Gerlach.CAM>
It's merged and backported. |
Part of #95913
This PR expands, edits and formats the CPython Bytecode Changes section in the What's New in Python 3.11 document.
On the editing side, like with the other PRs, it:
Likewise, it adds some missing additional opcodes that were added, replaced or removed in 3.11:
COPY_FREE_VARS,MAKE_CELL,PUSH_EXC_INFO,PREP_RE_RAISE_STAR,RETURN_GENERATOR,SENDandASYNC_GEN_WRAPSETUP_WITHandSETUP_ASYNC_WITHwithBEFORE_WITHPOP_BLOCK,SETUP_FINALLYandYIELD_FROMFinally, it consolidates the sets of replaced opcodes into a much quicker and easier to navigate table.
Rendered Preview
--> Rendered Preview (click to expand) <---