I'm assuming here that the goal is to make string building easier, better, and/or more discoverable, and that the io.StringIO discussion is just one way to achieve this. For example, I don't think (but maybe I'm wrong) that "must be a file-like object" is a goal here. If that's not the goal, then we should decide on the goal first. On 4/1/2020 5:43 PM, Paul Sokolovsky wrote:
On Wed, 1 Apr 2020 10:01:06 +0100 Paul Moore <p.f.moore@gmail.com> wrote:
I hope this isn't going to trigger another digression, but it seems to me that the answer is "nobody, unless they are taught about it, or work it out for themselves[1]". Roughly speaking, the answer would be about the same in idea as answers to the following questions:
* Who'd be using assignment expressions? (2nd way to do assignment, whoa!) * Who'd be using f-strings? (3rd (or more) way to do string formatting, bhoa!) * Who'd be writing s = s.removeprefix("foo") instead of "if s.startswith("foo"): s = s[3:]" (PEP616)? * Who'd be using binary operator @ ? * Who'd be using using unary operator + ?
I think some of those are bad examples, and not in the same league as this change. For example, the walrus operator is not just another way to do assignment (as it existed pre 3.9). But be that as it may, it seems to me that if you want to make string building easier, better, and/or more discoverable, then add a StringBuilder class. It's definitely going to be way more discoverable than "use io.StringIO" or "look up the FAQ and use ''.join()". And it could conceivably be more performant than either of those, since it would have fewer constraints. You can start with: class StringBuilder: def __init__(self): self.strio = io.StringIO() def __iadd__(self, s): self.strio.write(s) return self def __str__(self): return self.strio.getvalue() And you can later write it in C, make it use ''.join() instead of io.StringIO, or whatever else makes sense. I don't think changing the io.StringIO interface is a good decision: it's supposed to look file-like. I suppose if we could wave a magic wand and have all existing file-like objects support __iadd__ I might feel differently, but we can't. Whether such a StringBuilder is suitable for the stdlib, and if so where it goes, are questions to be answered if this approach works out. Eric