David Bremner <david@tethera.net> writes: > One way to break this up into more bite sized pieces would be to first > create one or more tests that fail with current notmuch, and mark those > as broken. Right - for the moment I just wanted to post what I had for consideration. I didn't want to spend too much more time on the approach if was uninteresting/inappropriate. One simple place to start might be the included T570-normalization.sh. Though perhaps that should be "canonicalization"? > Can you explain why notmuch is the right place to do this, and not > Xapian? I know we talked back and forth about this, but I never really > got a solid sense of what the conclusion was. Is it just dependencies? I have no strong opinion there, but to do the work in Xapian will require a new release at a minimum, and likely new dependencies. And generally speaking, I suppose I have a suspicion that application needs with respect to encoding "detection", tokenization, stemming, stop words, synonyms, phrase detection, etc. may be domain specific and complex enough that Xapian won't want to try to accommodate the broad array of possibilities, at least not in its core library. Though it might try to handle some or all of that by providing suitable customizability (presumably via callbacks or subclassing or...). And since I'm new to Xapian, I'm not completely sure what's already available. > It seems plausible to specify UTF-8 input for the library, but what > about the CLI? It seems like the canonicalization operation increases > the chance of mangling user input in non-UTF-8 locales. Yes, the key question: what does notmuch intend? i.e. given a sequence of bytes, how will notmuch interpret them? I think we should decide that, and document it clearly somewhere. The commit message describes my understanding of how things currently work, and if/when I get time, I'd like to propose some related documentation updates (perhaps to notmuch-search-terms or notmuch-insert/new?). Oh, and if I do understand things correctly, notmuch may already stand a chance of mangling any bytes that aren't an invalid UTF-8 byte sequence, but also aren't actually in UTF-8 (excepting encodings that are a strict subset of UTF-8, like ASCII). For example (if I did this right), [0xd1 0xa1] is valid UTF-8, producing omega "ѡ", and also valid Latin-1, producing "Ñ¡". > I suppose some upgrade code to canonicalize all the terms? That sounds > pretty slow. Perhaps, or I suppose you could just document that older indexed data might not be canonicalized, and that you should reindex if that matters to you. Although I suppose anyone with affected characters might well want to reindex if the canonical form isn't the one people normally receive (which seemed possible). Hmm, another question -- for terms, does notmuch store ordinal positions, Unicode character offsets, input byte offsets, or...? Canonicalization will of course change the latter. I imagine it might be possible to traverse the index terms and just detect and merge those affected, but no idea if that would be reasonable. > I really didn't look at the code very closely, but there were a > surprising number of calls to talloc_free. But those kind of details can > wait. Right, I wasn't sure what the policies were, so in most cases, I just tried to release the data when it was no longer needed. Thanks -- Rob Browning rlb @defaultvalue.org and @debian.org GPG as of 2011-07-10 E6A9 DA3C C9FD 1FF8 C676 D2C4 C0F0 39E9 ED1B 597A GPG as of 2002-11-03 14DD 432F AE39 534D B592 F9A0 25C8 D377 8C7E 73A4