Mailman 3 October 2017 - Python-ideas

Add pop_callback to deque signature
by Michel Desmoulin 04 Oct '17

04 Oct '17

My initial proposal would have been to add a data structure to allow something like deque, but allowing cheaper random accesses. However I realized it was a use case more suitable for an external lib. Thinking about how I would implement such a new data structure, I imagined several possibilities, one of them would be to couple a deque with a dict or a list. However, sharing data between a deque and another data structure is hard because, while you can easily hook on the element going in (since you put them in yourself), there is no efficient way to get back an element on its way out. On lists or dicts, if you remove an element, you can pop() it and you get back the removed element. On deque, if you set a maxlen of 5 and add a 6th element, if you want to get the element that has been removed, you need to check if the maxlen has been reached, and if yes, get a reference to the first element, then add the new one. It's inconvenient and of course slower than it needs to be given that deque are quite fast. So my more modest and realistic proposal would be to add a callback on deque signature: collections.deque(iterable, maxlen, pop_callback) pop_callback would accept any callable, and call it everytime an element is removed from the deque, allowing third party libraries to then do whatever they need with it.

6 9

Re: [Python-ideas] allow overriding files used for the input builtin
by Wren Turkal 02 Oct '17

02 Oct '17

Steven, Sorry for not replying to your message due to my mailing list incompetence (configured for digest at first). I am replying to your message here: https://mail.python.org/pipermail/python-ideas/2017-September/047239.html I have not figured out the issue with readline/arrow keys on the overridden files. Consider this current iteration an RFC on the concept. I’m sure there are bugs in there. ☺ wt

1 0

Re: [Python-ideas] allow overriding files used for the input builtin
by Wren Turkal 02 Oct '17

02 Oct '17

Paul, Sorry for not replying to your message yet, I lost it in a well-intentioned try at digest delivery. Sorry about that, I am replying to this message: https://mail.python.org/pipermail/python-ideas/2017-September/047236.html I would guess that the context manager to override stdout does it globally. One of the reasons I like the params for input is that is doesn’t have to override the value globally by setting the module variable, which would affect other concurrent uses of input(). Thoughts? Thanks, wt

1 0

Re: [Python-ideas] allow overriding files used for the input builtin
by Wren Turkal 02 Oct '17

02 Oct '17

Adam, I just realized that I had not responded to you. I’m sorry for not replying to the message as this was one of the messages that came in digest form. I am referring to your message here: https://mail.python.org/pipermail/python-ideas/2017-September/047238.html I looked over your comment about how input() should be like print(), but there is an issue that I see with it. It seems print() allows overriding the file used for the output. It seems like input() should allow that for symmetry. FWIW, I think your idea about making some file “readline” method more like input() is intriguing. I think I like the modification to input() better as it makes more sense to me as I can already change the file used by input, but it’s done through side effects now (i.e. setting sys.stdin to a different value). I think that removing the need for the side effect would be a good outcome. I feel like it’d at lease be more obvious what someone is doing than when I look at current use cases that redirect the stdin. wt

1 0