| 1 |
Brian Harring wrote: |
| 2 |
|
| 3 |
> -----BEGIN PGP SIGNED MESSAGE----- |
| 4 |
> Hash: SHA1 |
| 5 |
> |
| 6 |
> |
| 7 |
>> The basic problem in searching is actually that it isn't implemented |
| 8 |
>> smartly |
| 9 |
>> in current portage. I have working (emerge -s like) code that is |
| 10 |
>> blazingly |
| 11 |
>> fast as it does not actually open all ebuilds. |
| 12 |
> Searching works off of the cache for the most part, if a cache entry is |
| 13 |
> stale, it's updated (eg the ebuild is opened and srced). |
| 14 |
> Unless you're not checking the cache and updating it as you proceed, |
| 15 |
> you're implementation ought to suffer the same limitation. |
| 16 |
> |
| 17 |
>> Doing description searching is |
| 18 |
>> impossible to do fast without some kind of cache. I don't think |
| 19 |
>> creating a |
| 20 |
>> reliable cache for that is going to be a priority, |
| 21 |
> There are 2 things that need to be done (in my books at least) to step |
| 22 |
> up the speed of a description search- |
| 23 |
> A) sql based cache backend, whether sqlite or mysql. |
| 24 |
|
| 25 |
If it comes to this, I definately vote for sqlite. It's much faster than |
| 26 |
mysql for embedded apps, and you don't have to worry about whether or not |
| 27 |
the server is running. |
| 28 |
|
| 29 |
|
| 30 |
> Either that, or |
| 31 |
> extend the flat cache to store the descriptions in a central index. |
| 32 |
|
| 33 |
I think this is a good idea. AFAIK, this is what FreeBSD does. Perhaps |
| 34 |
tinycdb could be used? (OCaml has a CDB module available too. You can |
| 35 |
get it here: http://bleu.west.spy.net/~dustin/projects/ocaml/index.xtp |
| 36 |
(BSD licence)) |
| 37 |
|
| 38 |
|
| 39 |
> B) alter the search description alg so that instead of stepping through |
| 40 |
> each entry getting the description, we just state "give me all packages |
| 41 |
> that have a description matching blar", and leave it up to the backend |
| 42 |
> to decide what is the most efficient way to search. With flat cache, |
| 43 |
> we'd still have to go file by file; w/ a sql variant, it could take |
| 44 |
> advantage of the appropriate syntax. |
| 45 |
|
| 46 |
I personally prefer well designed hash table schemes. This is part of a |
| 47 |
base operating system, so it needs to be extremely efficient. IMO, SQL |
| 48 |
databases tend to encourage laziness. This is not to say that they don't |
| 49 |
have their place, but I don't think they belong in a package management |
| 50 |
system. |
| 51 |
|
| 52 |
Still, it's a novel idea, I think. And offers a great deal of flexibility. |
| 53 |
|
| 54 |
|
| 55 |
-- |
| 56 |
Jesse Guardiani, Systems Administrator |
| 57 |
WingNET Internet Services, |
| 58 |
P.O. Box 2605 // Cleveland, TN 37320-2605 |
| 59 |
423-559-LINK (v) 423-559-5145 (f) |
| 60 |
http://www.wingnet.net |
| 61 |
|
| 62 |
|
| 63 |
|
| 64 |
-- |
| 65 |
gentoo-performance@g.o mailing list |
Archives