Issue 12184 - Implement advanced formatting support for std.uni.InversionList

[ghost]

https://issues.dlang.org/show_bug.cgi?id=12184

[ghost]

std.uni seems to use a lot of magical things like opDispatch so I couldn't figure out which script has only a few intervals, I picked Cyrillic almost randomly. Is there a script that has just a few intervals so it looks nicer in the documented unittest?

[monarchdodra]

Somewhat related, but do we have any flags to say "show base"? Seems like it would be useful here.

[monarchdodra]

Hum... this only takes into account the "type" falg, but discards everything else (say, padding and whatnot). I think: something along the lines of:

FormatSpec!char fmt2 = fmt;
put(sink, "[");
formatValue(sink, val.a, fmt);
put(sink, "..");
formatValue(sink, val.b, fmt2);
put(sink, ")");

This is more generic and robust (I think): In particular, it handles explicit width for those with OCD that want alignment. Unfortunately, it doesn't work so well for handling the display of the base :/

[ghost]

Sorry, I haven't worked with other fmt options yet so I managed to forget about these. I'll have a look at this tomorrow.

[monarchdodra]

Sorry, I haven't worked with other fmt options yet so I managed to forget about these.

Well, what I mean is that (ideally) you shouldn't have to be re-inventing an Fmt to "format string" decoder. Just use the fmt straight up.

[ghost]

What do you mean by "show base"?

[monarchdodra]

What do you mean by "show base"?

I meant that it is the fomat flag that should choose wether or not to display "0x". In any case, I found it, it's '#'. EG:

    writefln("Value: [%5x]", 23);
    writefln("Value: [%#5x]", 23);

Prints:

Value: [    5]
Value: [  0x5]

So, if I have an inversion list, and if you use my code, you should be able to get:

writefln("%#06X", InversionList(0x30, 0x5000));

prints:

[0x0030..0x5000)

Now... how to customise space or not around the .. ? Just kidding :D

[ghost]

What's sad is that:

writefln("%#x", 0xabcd);
writefln("%#X", 0xabcd);

Prints:

0xabcd
0XABCD

But there's no 0xABCD. IOW keep the small x. Is this a bug or deliberate behavior?

[ghost]

Thanks for letting me know about being able to reuse fmt. I didn't know about this, so I've added it to this page: http://wiki.dlang.org/Defining_custom_print_format_specifiers#Re-using_existing_formatting_tools

[monarchdodra]

What's sad is that:

I think it's a bug. I've never head of 0XABCD before.

[ghost]

In the meantime is this good to go? I admit I don't like the duplicated code block there.

[monarchdodra]

In the meantime is this good to go?

The only thing I have doubts about is the 0X test, since I think it's a bug. I don't know if it's better to leave it in, and fix it later, or comment it out completely.

I admit I don't like the duplicated code block there.

When you want to loop "with a separator", you can avoid duplicate code with a loop with customized control:

    if (!range.empty)
        for ( ; ; )  
        {    
            //Process body
            range.popFront();
            if (range.empty) return;
            //Process separator
        }    

Another option, is to add the toString on the Interval object itself. I wish I could say that there is then a "cool range formatting trick", such as:

formattedWrite(sink, "%(µs %)", range);

But the truth is that I have been unable to "feed" the fmt flags back into the loop with that...

In any case, the code can then becomes pretty simple (copied almost verbatim from format):

            if (!val.empty)
            {
                formatElement(sink, range.front, fmt);
                range.popFront();
                foreach (i; range)
                {
                    put(w, " ");
                    formatElement(sink, i, fmt)
                }
            }

[ghost]

Updated.

[monarchdodra]

LGTM overall, but this comment isn't quite accurate anymore. Something simpler along the lines of "The formating flag is applied individually to each value" should do.