[PATCH 0/2] Use ? in format specifiers

["S. Gilles" <sgilles@xxxxxxx>]

As discussed on IRC, I'd like to print/fmt things to some fixed width,
but I don't know the width at compile time, only at runtime. This allows
passing '?' as an argument to any format parameter. When processed (in
sbfmtv), a ? means "Pull the next thing out of ap, turn it into a
string, and use THAT as the argument". So you can write things like

    std.put("{w=?}\n", 99, "foo")
    std.put("{w=?}\n", some_int_var, "bar")

or even (since everything is string-ified and perhaps unparsed)

   std.put("{w=?}\n", "99", "foo")

You can't use installed formatters to parse the ?-params, though, so
something like

   std.put("{w=?}\n", `Some_complicated_structure(99,"a",`XY), "foo")

probably won't fly.

Possible objections:

 - It feels inefficient to convert 99 into "99", then parse it back as
   the number 99. This is the price we pay for params being a simple
   (byte[:], byte[:])[:], and I think it's a reasonable price.

 - The temps array (of constructed byte[:]s that we must free before
   returning) feels a bit hacky. As has also been discussed a little on
   IRC, this really feels like the place where `auto' could shine, if
   the semantics were tweaked a bit.

 - The only current known use-case for this is variable-width padding;
   perhaps it's not worth losing 1:1 {}:argument counting for that?

 - ?

S. Gilles (2):
  Convert lib/test/fmt.myr to use testr
  Allow formatting arguments from variables

 lib/std/fmt.myr      |  32 ++++++--
 lib/std/test/fmt.myr | 190 +++++++++++++++++++++++++------------------
 test/matchctup.myr   |  32 ++++++++
 3 files changed, 172 insertions(+), 82 deletions(-)
 create mode 100644 test/matchctup.myr

-- 
2.24.0