[Gambas-user] New features on the wiki

[Benoît Minisini]

Le 08/12/2014 20:04, Tobias Boege a écrit :
> On Mon, 08 Dec 2014, Beno?t Minisini wrote:
>> Le 06/12/2014 14:25, Tobias Boege a ?crit :
>>> On Thu, 13 Nov 2014, Tobias Boege wrote:
>>>>> No, the source tree may not beavailable on the wiki, only the *.info files.
>>>>>
>>>>> Why not adding an optional "*.help" file that would come with the *.info
>>>>> file?
>>>>>
>>>>> This file could have almost the same format as the *.info file, except
>>>>> that it includes only the help. Your script would have to generate that
>>>>> file - I don't know what it does exactly at the moment...
>>>>>
>>>>> The format is:
>>>>>
>>>>> #<class #1>
>>>>> <symbol #1>
>>>>> '<Help line #1>
>>>>> '<Help line #2>
>>>>> '...
>>>>> <symbol #2>
>>>>> ...
>>>>> #<class #2>
>>>>> ...
>>>>>
>>>>> Then either you keep your script, and you send me the *.help files so
>>>>> that I put them on the wiki server.
>>>>>
>>>>> Better would be having your script inside the source tree, and let it
>>>>> run automatically on the C/C++ sources with specific Makefile.am rules.
>>>>>
>>>>> That script can be written in Gambas. But then it will be run at "make
>>>>> install" stage, once the interpreter has been compiled.
>>>>>
>>>>> What do you think?
>>>>>
>>>>
>>>> Maybe I should rewrite it in Gambas. At this time it's 5 programs (in 4
>>>> different languages!) piping into one another. The first four preprocess
>>>> the C/C++ comments and the last is a really hackish C program which takes
>>>> these help lines and merges them to the right place in a given .info file.
>>>> When I wrote these, I was all about (development) speed. It may be cool to
>>>> have all that in one program (and one language only).
>>>>
>>>> The above format (and an extra file for it!) would be really good because
>>>> that's information the program can gather from the C/C++ source files alone.
>>>> No need to mess with the .info files then.
>>>>
>>>> Weekend...
>>>>
>>>
>>> Of course I meant this weekend, not the one 3 weeks ago...
>>>
>>> The scripts are in the c2help directory in the source tree root since #6712.
>>> Try them on gb.openssl:
>>>
>>>     $ c2help/c2help.sh gb.openssl 2>/dev/null
>>>     #Digest
>>>     List
>>>     ' Return a list of all digests present in the local OpenSSL crypto library.
>>>     _get
>>>     ' Return a virtual object representing a digest algorithm by giving its
>>>     ' name. Valid names can be looked up from Digest.List.
>>>     IsSupported
>>>     ' Check whether the named digest algorithm is valid.
>>>     #.Digest.Method
>>>     Hash
>>>     ' Hash the given string using this digest algorithm.
>>>     _call
>>>     ' A synonym for Hash.
>>>     #Cipher
>>>     List
>>>     ' Return a list of all ciphers present in the local OpenSSL crypto library.
>>>     [...]
>>>
>>> One script is a gbs3 one, so all that needs to take place at "make install",
>>> as you said. Apart from that I need bash (with read and exit builtins), gawk,
>>> grep, sed and the coreutils (cat, tr and echo) -- to give an extensive list.
>>> Shouldn't be too exotic. But in case someone doesn't have egrep, sed or gawk,
>>> the help generation stage should be omitted since it's optional. I see no
>>> need to reduce the dependencies on my side.
>>>
>>> Regards,
>>> Tobi
>>>
>>
>> The way you did is not really useful. You should have written a pure
>> Gambas program with no dependencies, so that it can be run at "make
>> install" stage automatically.
>>
>> You have introduce many dependencies, which means a lot of problems if
>> in the future Gambas is ported on a non-GNU system (Android, for example).
>>
>
> I didn't know; I committed the scripts I had lying around for a year or so,
> just replaced the program that merged help comments into .info files with
> the script that outputs .help files.
>
> To reinvent the gawk- and sed-wheels should be the complicated part here but
> it doesn't seem to be too hard...

Is it so hard to extract help comments? I have never thought about that, 
so I guess it should no be too hard to extract them with a Gambas program...

-- 
Benoît Minisini