[Python-ideas] Dict joining using + and +=
steve at pearwood.info
Tue Mar 5 05:21:35 EST 2019
On Mon, Mar 04, 2019 at 10:18:13PM -0800, Amber Yust wrote:
> Adding the + operator for dictionaries feels like it would be a mistake in
> that it offers at most sugar-y benefits, but introduces the significant
> drawback of making it easier to introduced unintended errors.
What sort of errors?
I know that some (mis-)features are "bug magnets" that encourage people
to write buggy code, but I don't see how this proposal is worse than
In one way it is better, since D + E returns a new dict, instead of
over-writing the data in D. Ask any functional programmer, and they'll
tell you that we should avoid side-effects.
> This would be
> the first instance of "addition" where the result can potentially
> lose/overwrite data (lists and strings both preserve the full extent of
> each operand; Counters include the full value from each operand, etc).
I don't see why this is relevant to addition. It doesn't even apply to
numeric addition! If I give you the result of an addition:
say, you can't tell what the operands were. And that's not even getting
into the intricicies of floating point addition, which can violate
``(a + b) + c`` is not necessarily equal to ``a + (b + c)``
``x*(a + b)`` is not necessarily equal to ``x*a + x*b``
even for well-behaved, numeric floats (not NANs or INFs).
> Combining dictionaries is fundamentally an operation that requires more
> than one piece of information, because there's no single well-defined way
> to combine a pair of them.
Indeed, But some ways are more useful than others.
> Off the top of my head, I can think of at least
> 2 different common options (replacement aka .update(), combination of
> values a la Counter). Neither of these is really a more valid "addition" of
That's why we have subclasses and operator overloading :-)
By far the most commonly requested behaviour for this is copy-and-
update (or merge, if you prefer). But subclasses are free to define it
as they will, including:
- add values, as Counter already does;
- raise an exception if there is a duplicate key;
- "first seen wins"
or anything else.
More information about the Python-ideas