Add std.fmt.formatDuration and std.fmt.duration

[jdknezek]

formatDuration works on a writer, and duration wraps a u64 to allow pleasant injection into format strings.

[LemonBoy]

d will print the fp value with full precision, is that wanted?
Also, {s} for string formats please.

[jdknezek]

It's probably reasonable to limit the max precision to 3 decimal places (i.e. down to the next unit). I don't think there's any way to do max precision with fmt, but I can div(floor(mul))

[LemonBoy]

You can set the precision parameter for that.

[jdknezek]

Unfortunately @as(f64, 1) formatted with {d:.3} comes out as 1.000, where I would prefer to omit unnecessary precision

[zigazeljko]

Is using floats here really necessary? Since ns_remaining is a fixed-point number, you can manually split it into integral and fractional parts, and print them with formatInt.

[jdknezek]

Is eliminating the floating point operation worth the extra complexity?

-// Floor to 3 decimal placesconstkunits=ns*1000/unit.ns;
-constunits=@intToFloat(f64, kunits) /1000.0;
-tryformatFloatDecimal(units, .{}, writer);
+tryformatInt(kunits/1000, 10, false, .{}, writer);
+// Write up to 3 decimal places+constfrac=kunits%1000;
+if (frac>0) {
+trywriter.writeByte('.');
+if (frac%10>0) {
+tryformatInt(frac, 10, false, .{ .fill='0', .width=3 }, writer);
+ } elseif (frac%100>0) {
+tryformatInt(frac/10, 10, false, .{ .fill='0', .width=2 }, writer);
+ } else {
+tryformatInt(frac/100, 10, false, .{ .fill='0', .width=1 }, writer);
+ }
+ }

Serious question, I don't often work in environments where one floating operation would be noticeable

[jdknezek]

Not sure how performance compares, but another option:

constfrac=kunits%1000;
if (frac>0) {
trywriter.writeByte('.');
varfrac_buf: [3]u8=undefined;
_=formatIntBuf(&frac_buf, frac, 10, false, .{ .fill='0', .width=3 });
tryformatBuf(frac_buf[0..std.mem.indexOfScalar(u8, &frac_buf, '0') orelsefrac_buf.len], .{}, writer);
}

[jdknezek]

Thank you for the feedback! I have eliminated the floating point operations.

[jdknezek]

I've addressed the first pass of comments - thanks for the feedback!

[jdknezek]

I've now eliminated the floating point operations.

[andrewrk]

Thanks @jdknezek! Follow-up proposal: #7752