[ragel-users] Re: Is there a way to generate non-static variables?

Manoj Rajagopalan ma... at nanorex.com
Fri Mar 28 19:07:24 UTC 2008


In that case would it be too bad to provide a constructor kind of 
function which the developer must explicitly call? But this way we'd be 
simply be moving data-space reclaimed (?) to instruction space to encode 
these initializations in addition to storing the immediates so this 
might not be a great idea :-)

-- Manoj


Adrian Thurston wrote:
> If the variables were not static they would need to be initialized from
> static variables. So short of loading the data from a file you can't
> avoid the static data.
> 
> -Adrian
> 
> Manoj Rajagopalan wrote:
>> Hi Adrian,
>>
>>     In C/C++, the variables generated are static. While I can write data 
>> for different machine instantiations in different functions, the 
>> static-ness makes the data allocated persistent. Instead, if there were 
>> an optional way of disabling this which could be used inside functions 
>> allocation would be on the function's stack and the total data segment 
>> size for the program would remain modest.
>>
>>    I am in the process of writing a unit test suite where I have one 
>> master Ragel file and am including parts of it for non-invasive testing 
>> in my test-suite CPP files. I have to perform a write data for each line 
>> that I am testing and then for more complicated machines that build on 
>> lines tested in these simpler tests.
>>
>>    These tests are part of a potentially enormous test suite for a heavy 
>> application with many components. I suspect statically-allocated global 
>> data requirements can become quite large though I haven't faced any 
>> problems yet. Please correct me if I am wrong.
>>
>>    If this idea is worthy of exploration then Ragel could provide a 
>> 'nostatic' option to the 'write data' statement which we developers 
>> would use in some scope inside which we'd like the data to be 
>> exclusively visible.
>>
>> Thanks,
>> Manoj
>>
> 
> > 



More information about the ragel-users mailing list