std.PriorityQueue: Convert to an 'unmanaged'-style API #21433
Conversation
Resolves: ziglang#21432. Removed the allocator field from the PriorityQueue struct and adapted functions to take an allocator where needed. Updated tests accordingly.
Removed the allocator field from the PriorityDequeue struct and adapted functions to take an allocator where needed. Updated tests accordingly.
lib/std/priority_dequeue.zig
Outdated
| pub fn deinit(self: Self) void { | ||
| self.allocator.free(self.items); | ||
| pub fn deinit(self: Self, allocator: std.mem.Allocator) void { | ||
| if (self.items.len > 0) { |
There was a problem hiding this comment.
This isn't strictly necessary since free checks for 0-length slices but it still seemed odd to me to intentionally free something that was not a result of a call to alloc so I added this check.
lib/std/priority_queue.zig
Outdated
| pub fn deinit(self: Self) void { | ||
| self.allocator.free(self.allocatedSlice()); | ||
| pub fn deinit(self: Self, allocator: std.mem.Allocator) void { | ||
| if (self.items.len > 0) { |
There was a problem hiding this comment.
This isn't strictly necessary since free checks for 0-length slices but it still seemed odd to me to intentionally free something that was not a result of a call to alloc so I added this check.
|
I added back the I also noticed that what PriorityQueue was doing with the PriorityQueue and PriorityDequeue still have a
The latter I think matches the rest of the standard library more, though as a user I have sometimes found it a little odd that I need to access fields directly for this information rather than calling some consistent function. What do y'all think the best move would be here? |
This is probably motivated by what ArrayList does and I think people generally agree (myself included) that it's preferred over the alternative because it means that your |
Okay, I put it in its own commit so it's easy to undo. I'll undo it. |
…city" This reverts commit 42981fe.
Co-authored-by: Andrew Kelley <[email protected]>
Resolves: #21432.
Removed the allocator field from the PriorityQueue struct and adapted functions to take an allocator where needed.
Updated tests accordingly.
If these changes look good, I'll make corresponding changes to PriorityDequeue and include them in a follow-up patch.