Subject: Getting a replacement for sprintf ?!
In-Reply-To: <B1034373-B330-4661-8EF9-4E64638EFCD2@info-igor.org>
References: <ef94acf8-bc29-f666-0b19-efdd505f10d9@virtuell-zuhause.de>
<B1034373-B330-4661-8EF9-4E64638EFCD2@info-igor.org>
Message-ID: <395f4aab-f076-1ff0-0193-0dea37b13f7b@virtuell-zuhause.de>

Am 25.02.2019 um 16:57 schrieb Scott Hannahs:

Scott:

> Having used the C format string primitives for a long time, I seem to
> be comfortable with them and found them to be extremely flexible. I
> think there there isn?t a limit on string lengths with sprintf other
> than the Igor built in limit.

Yes, exactly that I'm talking about (2400 characters in IP8).

> sprintf is an operation and not a function in Igor terminology, to
> handle the variety of different number and type of optional
> parameters.

Yes, sorry for the confusion. And exactly it's operation typecharacters makes it not-so-nice to use.

> Your example is simple but doesn?t allow the full flexibility of the
> C string format specifiers where resolution and digits can be
> specified dynamically in the format string itself.

The format specifiers are hidden in the TODO comments ;).
The full "{}" supports all of C's printf specifiers see here -> http://fmtlib.net/dev/syntax.html.

> I think one could write a function with variable parameters to replace sprintf and
> avoid the temporary string but haven?t found it worthwhile.

AFAIK Igor XOP function parameters are typed so you can not just emulate that sprintf does.
An alternative to my optional parameter approach would be something like VARARGS support in plain IP.

> Or having a sprintf function that returns a string (sprintff ???)
> could be a wish list for Igor 9?

Would be nice indeed.

Thomas

>> On Feb 24, 2019, at 1:17 PM, Thomas Braun <thomas.braun at virtuell-zuhause.de> wrote:
>>
>> Hi Igorians,
>>
>> since quite some time I'm starting to dislike the sprintf function.
>>
>> Main reason is that I need to create a temporary string to be able to work with it and
>> also that it can not be used in arbitrary code as e.g. %s is limited to x number of characters.
>>
>> Since quite some time I've been using fmtlib [1] in my C++ projects.
>> In short it allows you to write something like
>>
>> fmt::format("The answer is {}.", 42);
>>
>> and that returns a string with the contents ?The answer is 42.?.
>>
>> Now there is no use in trying to get that exact syntax working in Igor Pro.
>>
>> But one could get quite close. Here is a hacky hack example:
>>
>> Function Examples()
>>
>> print f("Hi {s_a}!", s_a = "HAL")
>> print f("Are you up to counting from 0 to {v_a}?", v_a = 10)
>> print f("Sure ;) Wanna see {w_a}?", w_a = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10})
>> print f("In this case have fun with {s_a} and {v_a}s.", v_a=0, s_a="ones")
>>
>> End
>>
>> Function/S f(format, [v_a, s_a, w_a])
>> string format
>> variable v_a
>> string s_a
>> WAVE w_a
>>
>> variable numDefaultParams, i
>> string str, buf, w_buf
>>
>> if(ParamIsDefault(v_a))
>> numDefaultParams += 1
>> endif
>>
>> if(ParamIsDefault(s_a))
>> numDefaultParams += 1
>> endif
>>
>> if(ParamIsDefault(w_a))
>> numDefaultParams += 1
>> endif
>>
>> str = format
>>
>> if(numDefaultParams == 0)
>> // TODO check that str does not contain {}
>> return str
>> endif
>>
>> if(!ParamIsDefault(v_a))
>> // TODO make %g configurable from {}
>> sprintf buf, "%g", v_a
>> str = ReplaceString("{v_a}", str, buf)
>> endif
>>
>> if(!ParamIsDefault(s_a))
>> // TODO make configurable
>> str = ReplaceString("{s_a}", str, s_a)
>> endif
>>
>> if(!ParamIsDefault(w_a))
>> // TODO make configurable and support dimensions larger than 1D
>> // and also other wave types
>> w_buf = "["
>> for(i = 0; i < DimSize(w_a, 0); i += 1)
>> sprintf buf, "%g", w_a[i]
>> w_buf += buf
>>
>> if(i != DimSize(w_a, 0) - 1)
>> w_buf += ", "
>> endif
>> endfor
>> w_buf += "]"
>>
>> str = ReplaceString("{w_a}", str, w_buf)
>> endif
>>
>> return str
>> End
>>
>> which gives
>>
>> ?examples()
>> Hi HAL!
>> Are you up to counting from 0 to 10?
>> Sure ;) Wanna see? -> [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
>> In this case have fun with ones and 0s.
>>
>> and this is already quite nice (I think) because it avoids both my main problems with sprintf.
>>
>> My main problem at the moment is that it does not scale with the number of characters per type. If we would support up to 8 characters after the "_" we have already
>>
>> ?print/D 26^8*3
>> 626481193728
>>
>> combinations and thus a dump approach would need that many optional parameters. And to be usable we would need a way to iterate over them in a way with has sub-linear complexity.
>>
>> And we would probably also hit a hard-coded limit of the number of optional parameters or maximum line length or ....
>>
>> Thoughts?
>>
>> Thomas
>>
>> [1]: https://github.com/fmtlib/fmt
>> _______________________________________________
>> Info-igor mailing list
>> Info-igor at lists.info-igor.org
>> http://lists.info-igor.org/listinfo.cgi/info-igor-info-igor.org
>
> _______________________________________________
> Info-igor mailing list
> Info-igor at lists.info-igor.org
> http://lists.info-igor.org/listinfo.cgi/info-igor-info-igor.org
>