#musicbrainz

You have 9 _windows_ open..so do you never close Opera?

only when osx insists that I have to reboot

then I open it again :P

and when I run out of space for tabs in a window, I open a new window

I just tend to keep them all in one window, but I hit around 350 tabs yesterday :(

what do you need the 100 tabs for? to replace bookmarks?

typically just I've opened tabs and not bothered to close them :P

well I'm usually in the middle of doing something

but you can't find anything in that many tabs

I'm more liable to have a mess of terminal windows.

each with tons of tabs.

unless opera has something like search in tab titles

I can find things! just not instantly

sounds like an O(n) operation

well I leave enough of the tab name visible so I just need to go through the windows quickly

the only time that fails is with mb edits, I can't remember which edit number was a particular edit :P

why I have about 15 tabs open in a single window, I'm getting lost

that's when I start bookmarking things I still need and close the tabs

I have just been on the phone to a lovely lady called Tina at DeAgostini UK. She is doing her best to get me a list of Orbis/DeAgostini relases/catalog numbers in electronic format

I don't know the back story with this, but be careful. Label metadata is often way more messy than you'd expect.

I know. But I would still say that it is a better source than ebay listings :-)

The series I'm researching was released in UK, Germany, Netherlands Italy & France (and others). In France it was Les Genies de Blues

Catalog numbers & sequences are fubar, and only DeAgostini will know for sure the significance of the diverse catalog numbers

I once saw a prototype browser that displayed a graphical tree of your tabs/history, allowing you to see instantly how you got to a particular page

luks hi, would you have any time to help a bit with some plugin issues over next day or 2? am trying but not getting where i want, getting somewhere though

RifRaf: if you ping me tommorow, I'll help

too busy to read/write any code at the moment

ok cool, need to get away from code for a bit here too i think

I cannot believe my luck. When I called DeAgostini, I spoke to probably the only person in the world who considered it important to document their releases - catalog numbers in each country, release dates, track listings, source of the tracks. All typed with hand-written notes. She is scanning it all and sending it to me. REckons it will take her a few days to scan

O.o

only downside is then you won't have the fun of hunting down cat#s and release dates :p

She even knows which release were taken from the the original french series, "Les Genies du Blues'

finally :) a working part number function

Munger: if you thought your regexp was nuts, try this one: http://musicbrainz.pastebin.com/mc455b11

WTF kind of insanity is that? :-D

that would be a functional PartNumberStyle routine :D

I think the word you were looking for was 'dysfunctional'

we allow way too much garbage to be passed in to GC :P

What does that regex actually do?

seriously brian, this is very close to completely unmaintainable code

3 || Part‐Handler, Parts 1, 2, 3 - all handled

luks: I know... yet it's also quite needed

I don't it

there is always a way to write readable code

*doubt

what do you find unmaintainable about it?

the HUGE regular expression, for example

"do everything in a single place"

it simplifies down quite a good bit, once you take out the Deseret hunk of it

maintainable code is usually "do only as little as possible in a single place"

and then you glue the pieces together

Munger: find me a good way to access a range of plane 1 unicode with JS in a regexp, I'll quite happily swap out the 44 specifics

If it's a search for this, replace with that regex, then a lookup table in a separate file is probaly better

you need to write a lot of small functions, not a single huge piece of code

the old code worked by using a lot of small functions. Did you really find it readable?

well, I can do nothing about that, but I'm sure I'm right

yes, I was able to fix it any time I needed

I don't know what exactly you are trying to achieve, but that regex makes me shudder. Spider sense is tingling

put it this way: I have yet to find a single problem result that I cannot pinpoint to an exact line of code, just using console.log and moving up or down in the code.

In words of one syllable, what does that code actually do?

the old GC code, on the other hand, was a total nightmare, "fixing" things by merely adding in yet another specific fix

I guarantee you that you or somebody else will find this code problematic for the reasons I said

Munger: it takes text, finds all "part" words, and does all it can to split part name from part number, then makes a valid part-style string from it

it's the same reason why people are moving from php-like style of coding websites to the MVC model

luks: I don't plan to go anywhere

Do you have a sample of the data it is working on?

munger: foo part 1 2 3

It handles 1-3 now?

read PartNumberStyle... it's a pita in its permissiveness

and what does it parse 'foo part 1 2 3' into?

Foo, Parts 1, 2 & 3

Oh, doh

does 'foo part 1 2 3' appear on it's own line on the input file?

but also "foo pt 1- 2, 5", etc

Munger: any potential user input in a form field... there is no "input file"

"Foo, Parts 1 & 2, & 5" as output, btw

Ther is always an input file, even if it is post data

and luks: I see nothing about a text-parser which is handled better by a MVC-model, vs a FSM-model

Can there be multiple records on a single form submission, or is it just one?

What I am asking is if you ever see 'foo part 1 2 3 bar part 5-6 7'

all I know is, any possible cat-corner case I have yet to find, using every single GC-bug *ever* (wiki and trac), plus every single raw text input from the .nfo's for 1000 torrents, it's taken me less than 5 minutes to spot the problem area, and less than 8 hours to fix

brianfreud: you might not plan to go anywhere, but you can't guarantee life has no plans for you :P

Munger: nope

brianfreud: do you have a test suite for this code?

no, not at the moment

but the GC code is, unlike the old one, a true engine - it could easily be hooked in to one, if someone wrote one

so if you fix one problem, you don't know if 5 old problems is not broken again

well, actually, yes, 99.5% yes

shouldn't we have a testsuite in place first, to make sure we don't regress from what we already have?

luks: quick question, does ngs include album/track aliases?

any particular thing is handled only by a single section of code.

nikki: not directly, but it has all functionality for trivially adding that

there is no recursion, no looping, etc. It's a true FSM processor.

nikki: but of course you have the master track -> track, album -> release mapping

so that's very similar to aliases

so any change to PartNumberStyle code *only* affects PartNumberStyle text. If it doesn't match that regexp, it doesn't change the text, period.

OK. so you are always dealing with a single record in the format <foo> <["part"|"pt"] > [range] (, [range])...

ah...

Munger: s?\,?[\s\[\(]p(?:ar)?ts?\.?\b\s , but yes

* \s?\,?[\s\[\(]p(?:ar)?ts?\.?\b\s

nikki: well, for example you have a release group "Foo" with releases "Foo" and "Bar"

so if you search for the release group, you can use both "Foo" and "Bar" as aliases

Whoa! Steady on with the regular expressions :-D

(part (pt (parts (pts (part. (pt. (parts. (pts. part pt parts pts part. pt. parts. pts. etc

nikki: I even had the idea to not have track/release group titles manually entered

but to derive them from release/"tracklist track" titles

luks: all I know is, the old GC code took weeks to figure out how it did *anything*. This GC code, at least to me, is 100% comprehensible immediately. You may not get quite how something works, off the bat, but you can always spot exactly where situation X is handled, and why results Y or Z occur.

It seems to me that you are better off scanning the input data for number ranges and expanding them into a list of all numbers, and then re-assembling them into your required output. e.g. 'foo part 1-3 5 6' expands to 'foo part 1',foo part 2', foo part 3', foo part 5'.'foo part 6'.

Munger: you're assuming that part #'s are digits

part a, parts a-b, part foo, part bar, etc

luks: hmm... would be nice if some things could be aliases without having another pseudo-release though, e.g. that "- Human" song which everyone tries to rename to "Human"

I am giong on the data you gave me, but no, I am assuming no such thing. You need to identify all the input part number and expand them into individual records however, so that you can manipulate them as you wish

Theres a song called " - Human"? Whats the deal with the dash?

hm

you mean aliases that would be usable by taggers?

If the user input 'foo part 1-3 2', do you catch the duplicate?

Munger: it's "Guess Case", not "Replacement for User's Brain" :D

I'm not sure to be honest... it's hard to draw a line between 'search hint' and 'alias for people to use'

it's already smarter than the old GC, as it is

I can't think of a way how to use track/release aliases for tagging

how you specify that you want to use this particular alias?

Ah! If you had told me you were working on the Guess Case code when I asked, I would have had a better idea of the problem. Now I can think clearly about it

:-)

ah, sorry, thought you knew :)

No, but my question is still valid. To properly clean it, you need to catch duplicates

yes, but it merely attempts to apply the guideline. It tries to catch stupidity, but catching that part piece 3 is within the range of part pieces 1 and 2... a bit more than I think it needs to try to do (more than it already does, at least)

whoever: the current GC code: http://musicbrainz.pastebin.com/m4b3775e5 and all known issues: http://brianfreud.info/gcissues.txt

I guess if it were me, I'd show some sort of has-alternate-names-indicator which people could click on to show a list of alternatives

heh what do you know, I have " - Human" lol

So if I understand this right, when parsing input records, some of them have the word 'Part', 'Pt', 'Pts' or whatever and 1 or more numbers following. You need to catch these patterns and normalize them?

yes

brianfreud: http://www.youtube.com/view_play_list?p=BDAB2BA... - really good videos on this topic

...or Parts a-c

think of the worst freedb data you've ever seen. now try to fix it.