[chaley]

Originally Posted by davidfor

chaley: I like it. My vote goes for the first version. The second version is a lot more flexible, but I think the first is much simpler to read and understand. With the second I know I would mess up the arguments and spend hours trying to sort out. And if I didn't put them in order, I would miss a range somewhere.

I sort of like the idea of having it as:

first_matching_compare(v, val1, lt1, val2, lt2,  ..., ltn, elseval)

That reads to me as: val1 if v<lt1, val2 if lt1<=v<lt2, val3 if lt2<=v<lt3 and so on until elseval if v>=ltn. But, I think that would be inconsistent with the other template functions.

You are right, that is inconsistent with all the other similar functions.

I submitted the changes to Kovid (not yet merged) for a "first_matching_cmp" function. I decided to use "cmp" so that it was consistent with the numeric comparision function "cmp" and not confused with strcmp. If there is demand I can also add a first_matching_strcmp, but I don't see many use cases for it.

The inline doc for the new function is

_('first_matching_cmp(val, cmp1, result1, cmp2, r2, ..., else_result) -- '
            'compares "val < cmpN" in sequence, returning resultN for '
            'the first comparison that succeeds. Returns else_result '
            'if no comparison succeeds. Example: '
            'first_matching_cmp(10,5,"small",10,"middle",15,"large","giant") '
            'returns "middle". The same example with a first value of 16 '
            'returns "giant".')

EDIT: Kovid committed the change to the "master" branch.