[silpa-discuss] Re: Optimization idea

[Vasudev Kamath]

Yeah that clarifies my all doubts so we can go ahead with your idea which will 
definitely improve the efficiency. We need to finalize the format for indexing 
file 
and we should have some way to create this file automatically by just giving 
the dictionary name as input. Let me research on this once the format of index 
file is finalized

Thanks and Regards
Vasudev Kamath
On Saturday 24 Apr 2010 10:22:22 pm Santhosh Thottingal wrote:
> On Sat, Apr 24, 2010 at 8:16 PM, Vasudev Kamath <address@hidden> 
wrote:
> > Hi,
> > Idea looks good but i've some doubts here.
> > 1. Index file is it based on first letters or language. What i meant is
> > we will have a single file with words for all languages in it and index
> > file contains starting line number for each languages or is it for each
> > letter in each language?.
> 
> Separate dictionaries for each languages. Why I don't want to touch
> the existing dictionary is , these dictionaries are coming from
> various upstreams projects , mainly hunspell_xx_IN dictionaries, where
> xx_IN is any Indian languages.
> 
> > 2. I looked at the linecache module, we have getline function which
> > allows us  to get a particualar line specifed by line number. So we may
> > need to loop to read all the words incrementing the line number
> 
> Yeah, I just now looked into the linecache modules code, It is just
> wrapper over getlines. But instead of using linecache,
> file.seek(position) should not have this limitation.
> 
> > 3. How we will know where the words starting with a letter comes to end.
> > i.e is it sufficient only to have starting line number in the index file
> > or do we need ending line number?. There is one way to calculate ending
> > line number since index file contains next letter's starting index we can
> > stop 1 line before that number but how about the processing complexity or
> > implementation strategy?.
> 
> We need to know only start position. we can continue reading lines
> till the first letter changes.
> 
> 
> Thanks
> Santhosh
>