Bug 21808 – std.format: It should be possible to change the order of key and value of AAs.

When formatting AAs, one can use %(..%|..%) where the left side formats key--value pairs using two format specifiers. Here, positional formatting of the form %1$..s does not do its job. Position arguments are ignored. Test case: import std.stdio; auto spelled = [ 1 : "one", 10 : "ten", 100 : "hundred" ]; // expected: prints, in any order, 1 (one), 10 (ten), 100 (hundred) // result: prints, in any order, 1 (one), 10 (ten), 100 (hundred) [ok] writefln("%-(%1$s (%2$s)%|, %)", spelled); // expected: prints, in any order, one (1), ten (10), hundred (100) // result: prints, in any order, 1 (one), 10 (ten), 100 (hundred) [fail] writefln("%-(%2$s (%1$s)%|, %)", spelled);

This is related to https://issues.dlang.org/show_bug.cgi?id=18599 but not identical.

Changing this to "enhancement", because normal use of positional parameters has a completely different meaning - it gives the position of an argument in the function call and has nothing to do with the order of the elements of a single argument. Anyway, I see the need for the possibility to change the order; and positional arguments don't make any sense inside of a compound specifier. Therefore I think, it is OK to (ab)use them like suggested here. But then it's not a bug, but an enhancement.

@berni44 created dlang/phobos pull request #8020 "Fix Issue 21808 - std.format: It should be possible to change the order of key and value of AAs." fixing this issue: - Fix Issue 21808 - std.format: It should be possible to change the order of key and value of AAs. https://github.com/dlang/phobos/pull/8020

(In reply to Berni44 from comment #2) > Changing this to "enhancement", because normal use of positional parameters > has a completely different meaning - it gives the position of an argument in > the function call and has nothing to do with the order of the elements of a > single argument. > > Anyway, I see the need for the possibility to change the order; and > positional arguments don't make any sense inside of a compound specifier. > Therefore I think, it is OK to (ab)use them like suggested here. But then > it's not a bug, but an enhancement. For tuples' %( %) expansion format, reordering works: import std.typecons, std.stdio; writefln("%( %2$s (%1$s) %)", tuple(1, "one")); // prints: one (1)

(In reply to Bolpat from comment #4) > For tuples' %( %) expansion format, reordering works: > > import std.typecons, std.stdio; > writefln("%( %2$s (%1$s) %)", tuple(1, "one")); > // prints: one (1) I'd guess, that this was unintended (there is no unittest containing this, nor was it mentioned in the PR introducing this) and should be considered a bug (or an unknown feature?). Compare: writefln("%( %2$s (%1$s) %|%)", tuple(1, "one")); Now it throws. The reason, why the first version works is, because under the hood it is transformed to something similar to writefln(" %2$s (%1$s) ", 1, "one"); And now you've got the elements of the tuple as arguments. Anyway, as I wrote above and in the PR: I like the idea. Let's see, if it will be accepted.

dlang/phobos pull request #8020 "Fix Issue 21808 - std.format: It should be possible to change the order of key and value of AAs." was merged into master: - fbc4c53d7bd1916d5253c6d4a7bd2d61b80d31c3 by berni44: Fix Issue 21808 - std.format: It should be possible to change the order of key and value of AAs. https://github.com/dlang/phobos/pull/8020