A Q concerning map (a comparison to Maple V)

[Tim Peters]

[Franz GEIGER]
> > map(replace, ["abc", "cde"], "c", "C");
> > instead of having to create a lambda function.
> > Wouldn't it make sense to have that in Python too? Do there come
> > more such opportunities into your mind?

[Peter Schneider-Kamp]
> If I get you right you intend this to mean:
>
>   map(replace, ["abc", "cde"], ["c"]*2, ["C"]*2)
>
> So basically your proposal is to reuse map parameters
> which are no sequences in every iteration over the
> parameters who are,

It's hard to tell for sure, but that was my guess too.  In array languages,
this kind of thing is ubiquitous and is often called "scalar broadcast"
(where a scalar is any atomic value (as opposed to an array), and is
"broadcast" to every position of the arrays in which it's combined via some
operation).

There was a long debate about this a few years ago on c.l.py, where I
championed scalar broadcast in map specifically.  It collapsed under its own
hideous weight when-- as such things always do <wink> --that originally
modest goal got hijacked by people seeking to extend the semantics to every
corner of the language.  That won't happen.

> ...
> But how would you decide if "abc" is a sequence parameter
> or not?

That is a problem!  Kinda.  Mostly people want to broadcast numbers (as in
Franz's original example), and there's no problem there.  A cheap workaround
for sequences that are desired to be treated as scalars would have been to
introduce a new builtin scalar() function, that simply hid the
"sequenceness" of its argument from map.

Note that the listcomps in 2.0 make scalar broadcast much easier to spell;
e.g.

    [replace(x, "c", "C") for x in ("abc", "cde")]
    [x**2 + 5 for x in L]

So the better solution is not to extend map, but to forget it <0.9 wink>.

a-life-strategy-of-universal-applicability-ly y'rs  - tim