port search

[Ryan Schmidt]

On Sep 23, 2007, at 14:50, Randall Wood wrote:

> On 23 Sep 2007, at 14:21, Kevin Walzer wrote:
>
>> Ryan Schmidt wrote:
>>
>>> So.... it sounds like people want this feature, so.... it sounds  
>>> like it should be put into base, rather than having each GUI have  
>>> to reimplement it. And if "port search" was too slow, and you  
>>> came up with something faster, why don't we put that in base as  
>>> well?
>>
>> PortAuthority's method works by running "port list" on startup,  
>> caching the data, then iterating through that when the user  
>> searches for a term. It's faster than "port search" only because  
>> it doesn't have to launch a new process of tclsh when the user  
>> searches for a term. This method make sense for a GUI because it  
>> reduces the number of times it has to call out to the command-line  
>> tool, but it doesn't represent any sort of improvement in the  
>> command-line tool itself. Imagine having port run "port list"  
>> every time it started up...the result would be slower, not faster.
>
> I think all the GUIs cache most of the port data if only because  
> they need to display it and display different views of it rapidly.  
> There is no use for having the port command allocate something on  
> the order of 5 megs of RAM and walk the tree just to run a faster  
> search. The GUIs have the luxury of long training - users kind of  
> expect it to take a little longer to start before running fast,  
> while CLIs need to be fast and lean.

Understood! I agree -- that's a nice optimization for a GUI that  
won't work for a CLI. Oh well. :)