hi, printf-like function in FASMLIB (called "text.write.format") almost done.

Format string looks just like one in libc, just more limited:
- any normal chars are printed as are
- char "\" is escape char, so following char is printed as-is. This is needed to display "%".
- char "%" means variable value.

Behind "%" can be:
- optional "0" to pad text with '0' instead of spaces
- optional decimal number to specify padding size
- type identifier

Problem for me is the type identifier. I need following types:
- constant string (just pointer to string)
- string (pointer + size of buffer)
- unsigned decimal dword
- signed decimal dword
- unsigned hex lowcase dword
- signed hex lowcase dword
- unsigned hex upcase dword
- signed hex upcase dword

what should be letters for them?
for now i would use
- "c" for constant string
- "s" for string
- "d" for decimal
- "h" for lowcase hex
- "H" for upcase hex

still, how to specify signed/unsigned? Should i go for default signed, and specify unsigned with "u" prefix (like "%ud", "%uh")?

I would personally prefer unsigned as default, but then i see problem how to specify signed, "s" is already used for strings

any ideas welcomed

maybe signed could be specified with "%-h"? Or other special char as prefix?

that sounds very interesting,
for the signed numbers, i like the "%u" way !

BTW do you plan to add unicode too ?
thx

flaith: so you like "%d" and "%h" to print signed numbers by default?

about unicode, i am not sure yet. If it is supported widely on linux, i might think about it.

Well i like it, but i'm not sure that's the right way.
I think it'll be better to use the "C" format

nope. in C you have "int", "short", "long", and no idea how big they are. completely unusable in asm.

I too would prefer the unsigned to be default.
And for specifying signed I would suggest the "i" letter (like in IMUL instruction): %id, %ih, etc.

allright, here is newest proposal:

% can be followed by:
- optional "0" to pad with zeroes
- optional decimal padding size (eg. "%15d", "%08h")
- optional "i" for signed (eg, "%id", "%ih")
- optional "." and decimal number to specify number of decimal places for "f" (eg. "%.4f", "%04.5f")
- optional "p" to say argument is POINTER to type (useful for bytes, words, qwords and tenbytes)

followed by optional size modifier:
- "b" as byte
- "w" as word
- "q" as qword (will not be supported yet)
- "t" as pointer to ten-byte (80bit, double-extended precision floating, not supported yet)
if none is specified, dword is default

followed by type:
- "d" for decimal number
- "h" or "x" for hexadecimal number
- "c" for char
- "s" for constant string (without buffer length)
- "S" for string buffer (2 dwords: string pointer, buffer length)
- "f" for floating point value as decimal (will not be supported yet)

what do you think about it? except "S" for string buffer, i like it

I think to avoid having to learn another sting format layout it would be a good idea to keep the existing 'C' style as much a possible by only adding functionality and not redefining existing escape sequences. This makes code movement from one language to the other easier.

So much code is already written in 'C' we cannot go too far off the beaten track else we will get confused when reading the different codes.

revolution: yes, i know this is a problem. But could you give me example how to keep up with C?

for example:
- Decimal qword should be "ld" or "lld"? And what does "l" or "ll" say to assembly programmer?
- How about dword / (pointed/double-pushed) qword / (pointed) tbyte floating point number?
- Do you think signed decimals should be default, and "u" used for unsigned?
- Do you think there is no need for signed hexadecimal output in assembly?

These were main problems. And I just couldn't work out any way to merge C syntax and requirements of assembly code.

revolution: yes, i know this is a problem. But could you give me example how to keep up with C?

for example:
- Decimal qword should be "ld" or "lld"? And what does "l" or "ll" say to assembly programmer?
- How about dword / (pointed/double-pushed) qword / (pointed) tbyte floating point number?
- Do you think signed decimals should be default, and "u" used for unsigned?
- Do you think there is no need for signed hexadecimal output in assembly?

These were main problems. And I just couldn't work out any way to merge C syntax and requirements of assembly code.