Title of the article is "How…", not "Why…", and "why" is not discussed. (My understanding is that the "why" is "because the implementation acted that way without an official guarantee, and folks depended on the implementation detail, so it became guaranteed.)
I was under that impression as well, but it was added as an implementation in 3.6 (https://bugs.python.org/issue27350), so I don't know that too many people would have become dependent on it before it was official in 3.7: https://docs.python.org/3/whatsnew/3.7.html
I think the "how/why" thing is just the automatic title de-clickbait-ifer going a little haywire?
I don't think this is the "why". What you're depicting is what happened in JavaScript. Dictionary keys in Python always had that (to me) annoying property that they preserved insertion order until they don't. I'd frankly much prefer if they'd always be iterated in random order each time they're traversed.
I'd love to see the results of mandating a random order dict impl at an actual company/org (but hate to be forced to participate). Hopefully you hired developers who really like to write sorting algos.
Swift (heavily used by Apple) has randomly ordered dictionaries for security:
> In particular, random seeding enables better protection against (accidental or deliberate) hash-flooding attacks
https://forums.swift.org/t/psa-the-stdlib-now-uses-randomly-...
Perhaps not unrelated to why Python is the #1 most popular language while Swift is #22 https://www.tiobe.com/tiobe-index/
Swift isn’t popular because its Dictionary type uses randomly ordered keys?
It certainly could be a reason among many. Just look at the thread GP shared, containing multiple years' worth of users voicing frustration at the introduction of this behavior.
Probably the inference is YAGNI .
Well, that is how hash tables in go works, so you'd not have to look that far.
Perl since 5.8.something has had the option of perturbing the hash function, so it is different from run to run. You can also set the set to a given value in order to lock in the sequence.
In any case, it is not ordered. If you want that, you have to explicitly sort the keys of the hash.
Great. Maybe GP will go a step farther and also mandate arrays that return elements in random order too. Relying on insertion order for any reason is for weaklings.
^ came here to say exactly this
Thanks, How makes much more sense. The post title is dumb.
Why is Hackernews news for hackers?
The URL for the post includes "why" (en-why-is-python-ordereddict-ordered), so I suppose the title of the article was updated after the HN post was created. The site's native language is Chinese and I'm guessing the post was translated via automation. In fact, if you run the article through an automated translation service (Google Translate), it reproduces the "Why" title.
Thanks, makes sense.
Wrong title, article has "How Does Python’s OrderedDict Maintain Order?"