Duplicate branch list to avoid skipping entries

[cottsay]

This loop is changing the list of branches while it's iterating over it. I can't explain why this hasn't been blowing up builds for the past decade, but at least on Python 3.14, this can result in skipping entries in the list when other entries are removed.

[christophebedard]

not to nitpick, but would .copy() convey the intention better?

[cottsay]

not to nitpick, but would .copy() convey the intention better?

It may convey intention better, yeah. That said I think it's a pretty standard practice. Looks like that was introduced in Python 3.3, which is relatively recent against the lifetime of Bloom, but we're only supporting back to Python 3.6 right now so it would be okay.

Not that duplicating our list of <20 strings would actually be a performance concern, but I did some light benchmarking and constructing a new (mutable) list using list.copy() was roughly 18% slower than creating a new (immutable) tuple (for a list of 6 strings). Pretty irrelevant here, but good to know.

[tfoote]

The copy indicator is definitely easier to read/maintain which is best for this use case.

Interesting that the tuple constructor is that much faster.

[christophebedard]

Not that duplicating our list of <20 strings would actually be a performance concern, but I did some light benchmarking and constructing a new (mutable) list using list.copy() was roughly 18% slower than creating a new (immutable) tuple (for a list of 6 strings). Pretty irrelevant here, but good to know.

Interesting!