Option to specify whether decimal integer fields are signed or unsigned

[dc42]

This is a great tool! I would like the option to specify whether a decimal value should be interpreted and displayed as a signed or an unsigned value. Currently it appears that a 16-bit decimal field is always interpreted as unsigned, whereas a 32-bit field is interpreted as signed. For example, this data:

258d2d2200000001000000000100000000ffffffffffff000000000000000000000000000000a7f38
hhhhhhhh hhhhhhhh hhhhhhhhhh hhhhhhhh 16d 32d 

displays:

258d2d22 00000001 0000000001 00000000 65535 00000000-1 00000000 00000000 0000000000 0000a7f3 08

[zuckschwerdt]

Thanks. Yes we need better output formats someday. The decimal part relies on JS interpretation of rules to apply, the limit here is 52 bit for an integer then.
But getting a 2's complement for 32d really must be a bug. I'll take a look.

[zuckschwerdt]

The code to rework is ("it's unsafe to output more than 32 bits in one call" -- we mostly knew this would crash and burn)

bitbench/src/bitstring.js

Lines 106 to 122 in 457c32f

	// note: it's unsafe to output more than 32 bits in one call
	function formatBitsChar(bits, base = 16) {
	var num = 0
	var pad = Math.ceil(bits.length / Math.log2(base))
	while (bits.length)
	num = (num << 1) + bits.shift()
	if (base == 256) { // special case ascii
	var inv = num >= 128
	num %= 128
	var ctrl = (num < 32 || num == 127)
	if (ctrl)
	num = (num + 64) % 128
	var cls = (ctrl ? 'ctrl' : '') + (inv ? ' inv' : '')
	return `<span class="${cls}">${ctrl ? '^' : ''}${String.fromCharCode(num)}</span>`
	}
	return num.toString(base).padStart(pad, '0')
	}

[zuckschwerdt]

I'll check if this can be fixed using BigInt and then introduce s as sign-decimal format specifier.

[zuckschwerdt]

Numbers bigger than 31 bit are now fixed with 4df41e6 using multiplication by 2 instead of left-shift which is signed modulo 32.
Also numbers bigger than the maximum integer safe 52 bits use BigInt.

[zuckschwerdt]

A signed decimal format (2's complement, only up to 32 bit width) has been added as "s" with 464fa1e
Example in this BitBench.

[zuckschwerdt]

We could want signed for larger field width, and for e.g. hex.
Larger width might be possible if we apply the 2's complement ourselfs -- but probably needed step up from 32 bit is 64 bit which then also needs BigInt.
Other formats (hex) should be possible but will be a mess in the code.
Let's wait until someone really requires this :)

[dc42]

Wow, that was quick! Looks good to me. Might be worth mentioning in the docs that if you want signed little endian interpretation, the < character must come between the s and the number of bits. So s<32d works but <s32d doesn't.

I don't have any need to display signed hex values, and I think that would be an unusual thing to do. OTOH, interpreting 32-bit or 64-bit strings as IEEE floating point values may be needed some day.

[zuckschwerdt]

Oh, sorry, I thought about making signed a general modifier, e.g. $ but it's a format specifier right now, you'll want <32s

I'll see about e.g. 32f and 64f (also <32f and <64f).

[dc42]

Ah OK, 's' replaces 'd' instead of being a modifier. That works!

[zuckschwerdt]

Float is now supported, an example in this BitBench.