Introduce sizehint!(s, n; shrink = true) to controll shrinkage #51929
Conversation
petvana
added
collections
|
LGTM. Thanks for working on this. |
Co-authored-by: Jameson Nash <vtjnash@gmail.com>
Co-authored-by: Jameson Nash <vtjnash@gmail.com>
…lia into pv/dict-controll-shrinkage-v3
|
Are CI problems related? I'll update branch to test. |
|
CI passes and PR seems ready. |
|
Rebased #51903 on top of this now. Thanks! |
| @@ -1462,10 +1462,15 @@ For types that support `sizehint!`, | |||
| `Base`. | |||
|
|
|||
| 3. `empty!` is nearly costless (and O(1)) for types that support this kind of preallocation. | |||
|
|
|||
| 4. `shrink` controls if the collection can be shrunk. | |||
There was a problem hiding this comment.
Reading this, I don't understand what the argument does. Would you mind adding a more detailed explanation? Also it would seem more appropriate to mention the argument above, outside of the "# Notes on the performance model", as I wouldn't expect to have to read a performance section to find out the meaning of an argument.
There was a problem hiding this comment.
Thanks, this seems like a fair point. The new optional argument enables us to suggest not to shrink the already reserved capacity when suggesting new size by setting shrink = false. This is helpful for algorithms like merging two collection where you want to pre-allocate memory, but there is not reason to change (shrink) size if all the data fits. I'm not a native speaker, would you be able to propose a more fitting description?
There was a problem hiding this comment.
I've not a native speaker either, but see #52226. I hope I got it right.
This is a second part of #51629 to introduce
shrink = trueargument forsizehint!into public API. Fell free to propose a different name, or way to control shrinkage.Thinks to do:
VectorSetVectorSet