| jelmer | lifeless: so, I've been pondering doing a unity lens that uses bzr-search | 03:17 |
|---|---|---|
| jelmer | lifeless: a problem with that appears to be that currently all indexes are scattered across different branches; have you considered a global cache, or perhaps meta cache? | 03:17 |
| lifeless | jelmer: hiya | 04:24 |
| jelmer | lifeless: hi | 04:24 |
| lifeless | jelmer: needs a fast reachability oracle | 04:25 |
| lifeless | the basic form of a search is: load the metadata for all specified terms; choose the most selective (least occurences), do a bitmap AND of -all- the hits in that index with the next least selective, rinse and repeat | 04:27 |
| lifeless | so there are two issues with indices that cover more than the corpus for which hits are relevant | 04:27 |
| lifeless | firstly, a term that is very rare in the chosen branch may be common in another, making the search slower than it would be with the current approach | 04:28 |
| lifeless | secondly, we need to start filtering the index contents for revisions in the branch | 04:29 |
| lifeless | one way to do this, is to index all the revisions as terms of the branch (url) or branch (tip) or ... | 04:29 |
| lifeless | then have an implicit term added to all searches of the branch (url or tip or whatnot) | 04:29 |
| lifeless | there is of course a different but also interesting search 'what branches match this search' | 04:30 |
| jelmer | lifeless: ah, I see | 04:31 |
| jelmer | lifeless: thanks | 04:32 |
| lifeless | OR searches obviously need a more complex planner, I haven't written support for them yet | 04:32 |
| lifeless | I'd be delighted to discuss approaches if you want to hack in this area | 04:32 |
| lifeless | its pretty fun stuff | 04:32 |
| jelmer | I would be interested in hacking on and discussing this stuff some more some time | 04:32 |
| jelmer | about to fall asleep though, so perhaps some other time | 04:33 |
| lifeless | cool | 04:33 |
| lifeless | sleep well! | 04:33 |
| jelmer | thanks for the explanation so far | 04:33 |
| jelmer | thanks :) have a good day | 04:33 |
| lifeless | I shall ... well evening :) | 04:34 |
| lifeless | I have laptop, warm weather, beer. | 04:35 |
| === astralja1a is now known as astraljava | ||
| Movix | hello evereybody | 12:05 |
| Movix | i'm new to bazaar and did not found any way to do what i'm looking for. | 12:07 |
| Movix | I searched forums, tutorials and the online help and tried a lot the last days, but no chance. | 12:07 |
| Movix | so i wonder if this could be the right place to ask some questions or if it's better to post on a forum. | 12:07 |
| Movix | btw: i'am also new to VCS | 12:08 |
| === Quintasan_ is now known as Quintasan | ||
| LpSolit | I have a question about Loggerhead: I would like it to list all changes between two revid (~700 changes), but it always displays 20 changes at once only. How can we force it to display the exact list I want? | 13:04 |
| jelmer | LpSolit: I'm not sure if that is yet possible at the moment | 13:22 |
| LpSolit | oh, ok :( | 13:22 |
| jelmer | LpSolit: it shouldn't be too hard to patch it to do so though | 13:22 |
| LpSolit | I wanted to generate a list with all changes made between two consecutive releases of our product | 13:22 |
| === yofel_ is now known as yofel | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!