Hacker News Reader

Comment by someone13 | Hacker News Reader

Another method that I find isn't used enough: dict.setdefault. Instead of doing this:

    if 'key' in d:
        val = d['key']
    else:
        val = d['key'] = 'some default value'

You can do this:

    val = d.setdefault('key', 'some default value')

This is super-useful for configuration in frameworks, for example - it's a one-liner to set a default config value if the user didn't provide one.

View on HN

I find I use it most in multi-level data structures, not unlike Perl's auto-vivification system:

    logins = {}
    for user, date in getLogins():
      logins.setdefault(user, []).append(date)

Here you login ends up having one key per "user" with a list of "date".

Probably better done with a defaultdict:

  from collections import defaultdict
  logins = defaultdict(list)
  for user, date in getLogins():
    logins[user].append(date)

if you have a lot of them, plain dict may prove faster.

Not likely. Proof?

I'd agree if you'd said "not a lot of key/values," as the setup could supersede the subsequent speedups. However, in that case, the difference is negligible, and we should do what is "pythonic."

Yes, I meant a myriad of very small dicts. In this case a plain built-in is faster.

I think you might have glossed over another very useful "shortcut" not covered in this article, the use of multiple assignments in a single statement.:

    val = d['key'] = 'some default value'

Reply on news.ycombinator.com