12:59 AM
For numpy/scipy users: I've finally managed to build scipy and numpy on a windows machine.
For others, nevermind, just some nerdy thing has happened.
They are built against OpenBLAS instead of MKL using mingw-w64 compilers.

1:54 AM
@egreg The linguistics department at MIT was, for most of its time (until it moved to a crazy looking Frank Gehry building) in Building 20, a 'temporary' wooden building constructed during the Second World War and only demolished in 1998. :)
@percusse Surely this should be aimed at Windows users not numpy/scipy users. If you're using numpy/scipy on a Mac or Linux you presumably don't give a s*it. :)

2:49 AM
@DavidCarlisle Wow! With the Italian notation: 414.414 it would be even better, the comma is not symmetrical.

4 hours later…
6:19 AM
What would cause \box\cclv to have a negative height?
could that happen from overloading the insert mechanism?

@AGoldMan Adding boxes inside a box should never produce a negative height; a box can be assigned a negative height only by direct assignment to its \ht.

6:38 AM
@egreg does the standard LaTeX output routine make any such adjustments? I keep getting overprints with my footnotes b/c one of my pages has the main output box with its height as a negative number

@percusse mkl doesn't work?
@CarLaTeX yes it looks more symmetrical but it looks as wrong as a pineapple pizza, rep points should be integers no a little less that 414 and a half.

@DavidCarlisle I knew you would comment something of that kind :P

24 hours ago, by Harald Hanche-Olsen
Oct 11 at 16:46, by David Carlisle
Sep 10 at 12:03, by David Carlisle
@HaraldHanche-Olsen do you ever get a feeling of déjà vu in this chat room?

7:00 AM
@DavidCarlisle 414 414 is best. (Though the thin space looks awfully thin in this font.)

@HaraldHanche-Olsen we could put in a feature request for the new top bar design. I'm sure the stackexchange staff understand that displaying palindromes is an important, perhaps most important, feature of the top bar.
5

@DavidCarlisle :) And then you can start work towards 808 808.

@AGoldMan no, although any number of packages do any number of things to the output routine

@DavidCarlisle Though 600 009 (and several others like it) is a nice twist on the palindrome theme.
Sep 10 at 12:03, by Harald Hanche-Olsen
@DavidCarlisle Yeah, all the time, all over again.

7:16 AM
@DavidCarlisle LOL

yeah, it looks like it's dblfnote's fault

8:20 AM
Quack!

@PauloCereda Qua qua qua!

@CarLaTeX ooh an Italian duck

@PauloCereda <3

@CarLaTeX <3
Speaking of Italy, we have the GuIT meeting next Saturday!

@PauloCereda Of course!

8:37 AM
@CarLaTeX yaaaay

9:16 AM
@PauloCereda In the beautiful Mestre. ;-)

@egreg With Arrivederci Roma playing in the background. :)
@CarLaTeX ^^

@PauloCereda Nineta, monta in gondola, perhaps!

@egreg :D
@egreg: I should have used ballo del qua qua :)
With @CarLaTeX's friend Romana PowerPoint

9:33 AM
@PauloCereda You have been warned! :)

@CarLaTeX No one will notice it. :)

9:52 AM
@PauloCereda if we exclude the ones who was born or live in Italy

@CarLaTeX oh

10:27 AM
@DavidCarlisle MKL is possible only with proprietary intel fortran compiler.

@percusse ah yes I think I may have known that once, (of course we have that here, as well as our own compiler:-) is that currently the only public numpy build on windows?

@DavidCarlisle For win users, Christoph Gohlke is doing an incredible one-man-show lfd.uci.edu/~gohlke/pythonlibs providing windows builts with MKL. But since NumPy 1.14 and SciPy 1.0, guys figured out the toolchain needed to build windows wheels. Hence it's gonna be possible to do pip install though with OpenBLAS. But benchmarks show that it is not that bad against MKL

@PauloCereda go away and learn some atlas autocode

@PauloCereda I have the feeling a COBOL class would look more vivid...

10:38 AM
@DavidCarlisle you are mean
@TeXnician not wearing a tie while writing COBOL code is a syntactic error. :)

@PauloCereda
Jan 12 '16 at 12:15, by David Carlisle
@PauloCereda finished your thesis?

@percusse oh no

@percusse thanks I may ping you about this in a month or two:-)

@DavidCarlisle phew, for a minute I thought you were referring to the thesis message...

@PauloCereda :)

10:45 AM
@percusse ooh congrats sir

Oops edited the wrong message
@DavidCarlisle Yes sure. I've become a member of the SciPy team last month.

@percusse yay
@percusse: will you come to TUG 2018?

@PauloCereda where is it?

Jan 12 '16 at 12:15, by David Carlisle
@PauloCereda finished your thesis?

@percusse Rio de Janeiro
@DavidCarlisle oh no

10:48 AM
@percusse a long way away

@PauloCereda whoa, does TeX even compile over there?
3

@percusse after some beach, football and perhaps a caipirinha. :)

Missing an \item because it's too hot in here. I've lost \count
4

11:31 AM
@percusse ! Missing number, treated as zero.

! Missing wallet, report to the authorities.
Welcome to Brazil. :(

12:08 PM
@PauloCereda ouch!

@DavidCarlisle Don't tell this to egreg! — CarLaTeX 1 hour ago
@CarLaTeX OK I won't tell @egreg

@HaraldHanche-Olsen :(
@DavidCarlisle Who? @egreg?

@PauloCereda I am sure @egreg won't hear us if we keep our voices down.

@HaraldHanche-Olsen :)

12:17 PM
Any exciting news here?

@egreg No, nothing to see here, just move right along.

Would anyone here care to try out my two column paragraph footnotes package?

@DavidCarlisle -- oh, sure -- why not? tex.sx is certainly mentioned, even recommended, in tugboat, so suggestions on how to enjoy one's stay there are not unreasonable. (and, after all, it would only enhance your reputation as a bit of a prankster.)

12:40 PM
@barbarabeeton Eventually I've found the adequate translation for "simpatico burlone"!

12:50 PM
Don't use footnotes in your books, Don.
\author JILL ^{KNUTH} (1962)

@DavidCarlisle, @egreg Appreciate thoughts/input on comments on tex.stackexchange.com/questions/396567/…

Hi everybody.
Does anybody know if this is expected behaviour?
With babel hyphenrules seems to change \langname as well, even though the documentation says that should not happen.
\documentclass{article}
\usepackage[british]{babel}

\begin{document}
\begin{hyphenrules}{german}
\languagename
\end{hyphenrules}

\begin{hyphenrules}{british}
\languagename
\end{hyphenrules}

\begin{hyphenrules}{french}
\languagename
\end{hyphenrules}
\end{document}
From reading the documentation I'd expect the MWE to output 'british' three times, but I see 'german', 'british' and 'french'.
The docs say hyphenrules The environment hyphenrules can be used to select just the hyphenation rules. This
environment does not change \languagename and when the hyphenation rules
specified were not loaded it has no effect.

@moewe I don't see this in my version of babel.pdf.

@UlrikeFischer Mine (3.14) says that in the code explanations on p. 102

@moewe Ah. the code section. Well obviously the description is wrong, the code does \edef\languagename{#1}.

1:04 PM
@UlrikeFischer My hope was that the code should do something else and the explanation is right.
Ideally hyphenrules should not change \languagename as other packages use this to set up their multilingual features. hyphenrules should only change the hyphenation patterns.

@JosephWright yes saw that briefly although I see more comments since:-) will try later (but not sure I have anything to add)

1:18 PM
@moewe Well I'm not sure. Imho the core of language changes is the hyphenation patterns. Everything else (fontencoding, fix words etc) are more or less optional, so it would be odd if it wouldn't set \languagename. But I understand your need. Imho what you need is a command that holds the last language set by \selectlanguage/\otherlanguage and another with the language of a \foreignlanguage/\otherlanguage* call.

@UlrikeFischer Indeed.
But the problem is that csquotes and biblatex rely on \languagename.
And I suspect other packages also use this command to find out about the active language, not the hyphenation.

1:35 PM
@moewe Yes, but probably in the long run, it would help this packages too, if there where a clear and documented set \selectlanguagename \foreignlanguagename \hyphenlanguagename? . So I suggest a feature request ...

@UlrikeFischer You are right I assume, but I'm not entirely sure what I should request (and where). If there are too many commands like this, it may not always be clear which one is best for the job (\selectlanguagename vs. \foreignlanguagename)

1:52 PM
@JosephWright I can't understand what approach EA is having to the matter. It seems he wants to coerce TeX into being C or something. Good luck.

@egreg: @DavidCarlisle, @HaraldHanche-Olsen and I were supposed to not tell you something that @CarLaTeX has mentioned. We've been doing a pretty good job so far!
2

@PauloCereda My spies have already reported. :-P

@egreg ooh spies

@PauloCereda Intelligence agents, to be more precise. :-)

@egreg ooh we might need to summon the Scotland Yard.
2
@DavidCarlisle ^^

2:17 PM
@PauloCereda and you haven't seen this one: tex.stackexchange.com/posts/396548/revisions

@CarLaTeX ooh
@CarLaTeX Sadly not as famous as: github.com/davidcarlisle/dpctex/commit/…

@PauloCereda I know :):):)

@CarLaTeX ooh :D

@PauloCereda LOL

@CarLaTeX There's another missing %

2:24 PM
@egreg After \tikz{?

@CarLaTeX Look closely at my edit
@CarLaTeX Inside \tikz{...} space are ignored, but not after the final }.

@egreg Oooh I forgot that! And I asked a question about it some time ago!

:4059607 what does it claim we wrote?
Jan 12 '16 at 12:15, by David Carlisle
@PauloCereda finished your thesis?

@DavidCarlisle oh no
@DavidCarlisle even the system doesn't know. :)

2:39 PM
@DavidCarlisle @PauloCereda's thesis of course (hence an unknown title) ;)

@TeXnician oh

@PauloCereda I repeat my suggestion of employing a spam filter

2:50 PM
My battery's dead, see you after dinner!

@CarLaTeX Dish of the day: pineapple pizza?

is the a boxing duck ?
there are ducks of every profession it seems

@TeXnician No, orecchiette con friarielli :)

@alfred ooh boxing

@egreg I know: it's all very odd

3:08 PM
I got institutional access to the Wiley book repository. I am disappointed. :(

@PauloCereda ?

@alfred Mathematician's ducks

@JosephWright I thought the catalogue would have more books on theory.

19, and I hate them all

3:14 PM
@egreg High-school maths exercise?

@TeXnician That's from one of my exams. :-P

@egreg none: real ducks are not made of rubber

3:30 PM
@DavidCarlisle I count “19 images of a rubber duck”, to be more mathematically rigorous.

3:41 PM
@DavidCarlisle I am getting a pattern conflict for a list xetex --ini doesnt object to. It shouldnt do it right ?

I see Frank is getting to grips with Git

4:03 PM
@JosephWright poor Frank. :)
You threw him to the kraken, you naughty duck!

@PauloCereda That's Will: Frank has SourceTree

@JosephWright Ah I know the latter. :)

@JosephWright and l3build:-)
@JosephWright I tried installing kraken but it made this big window thing and I couldn't see where to type git commands, so I reverted to using xterm.

@DavidCarlisle ooh

@alfred I would suspect user error

4:09 PM
@DavidCarlisle :)
@DavidCarlisle Yup

@davi

@DavidCarlisle: wouldn't magit help you?

@DavidCarlisle OK. I'll repeat the operation.

@PauloCereda dunno what does it do (I saw it mentioned somewhere) but I just installled git kraken as Will mentioned it and I wondered if the windows application would work using cygwin shell at the back, it does seem to. (unlike some other tools I could mention if I was mean)

@DavidCarlisle :)
@DavidCarlisle: It's been a long while since my last Windows interaction. It seems that even MSYS now relies on a package manager (namely pacman, probably from Gentoo)

4:13 PM
@DavidCarlisle No I tried zja1 and zja8 on the list and xetex doesnt object

4:24 PM
same result. lualatex complains a conflist and xetex --ini doesnt catch it
conflict*

@alfred odd still it clearly is a conflict so one up to luatex:-)

I would prefer the opposite to be frank

@alfred I would prefer 1=0 (it makes lots of proofs easier) but sadly that's not how it is.
@alfred what would you want it to mean???

I guess I would like a way to know what LuaLaTex finds to be a conflict

@alfred it's obviously a conflict as it's zja in both cases, presumably if xetex doesn't error it allows the case where the numbers are in the same slot and just discards zja1

4:30 PM
it's not repetitions, i tested that

@DavidCarlisle ooh I like this thinking

@alfred ?? it is zja repeated

I had tried zja1 zja1 earleir
so what would constitute a conflict 1zja zja1 ? as we saw the first time

@alfred we seem to be going in circles. a pattern zja1 zja8 tries to assign both 1 and 8 to the end of any substring zja an that is a conflict.
@alfred there you can not resolve by just discarding the lower number you have to merge the patterns to make 1zja1

my zz.tex file now only contains 1oj oj1 \end and xetex --ini zz.tex still doesnt give an error

4:39 PM
@alfred that is not a reproducible test file (or if you mean you literally have that then you are not using \patterns at all.)

??

@alfred this is a reproducible test file:
\catcode\{=1
\catcode\}=2
\patterns{
1oj oj1
}
\end
which makes:
! Duplicate pattern.
l.6 1oj oj1

wait, you are saying I should have \patterns{} there ? Joking, completely my fault

@alfred xetex also gives the duplicate pattern error fo the zja1 zja8 example that you mentioned earlier.

all makes sense now

4:46 PM
@alfred how do you expect xetex to guess you are doing anything to do with patterns ?

somehow i had memorized that zz contained only the patterns, missing \patterns
I dont know exactly what --ini does
tried googling but ..
sorry about that

@alfred I showed you what you need before: you need catcode and lccode settings for {} and all non-ascii characters and you need \patterns and \end
@alfred google for initex (which is what the texbook calls it) recent distributions do not have a separate initex program they have tex --ini but it is the same thing

yes I know, I deleted parts of them inadvertently
I have now :
\catcode\{=1
\catcode\}=2
\catcode\ë=11 \lccode\ë=\ë
\catcode\ç=11 \lccode\ç=\ç

\patterns{gjer1 gjer8gj etc}

\end

5:07 PM
Fedora 27 final freeze!

is this right for ç' \catcode\ç=11 \lccode\ç=\ç \catcode\'=12 \lccode\'=\'

5:58 PM
@alfred yes although why discouraging and encouraging hyphen after gjer ?

6:32 PM
meow?

@PauloCereda you can't eat cats, ducks are more use

@DavidCarlisle oh no
That reminds me of Alf, was him famous on British TV?
He was very famous in Brazil.

@PauloCereda alf who?

@DavidCarlisle Hm it was an American sitcom:
ALF is an American sitcom that aired on NBC from September 22, 1986, to March 24, 1990. The title character is Gordon Shumway, a friendly extraterrestrial nicknamed ALF (an acronym for Alien Life Form), who crash lands in the garage of the suburban middle-class Tanner family. The series stars Max Wright as father Willie Tanner, Anne Schedeen as mother Kate Tanner, and Andrea Elson and Benji Gregory as their children, Lynn and Brian Tanner. ALF was performed by puppeteer Paul Fusco, who co-created the show with Tom Patchett. Produced by Alien Productions, ALF originally ran for four seasons and...

@PauloCereda don't recognise that description, although I had no TV for most of that period:-)

6:42 PM
@DavidCarlisle oh :)
The show was so bad it was so good. :)
@AlanMunn: do you remember Alf? :)

@PauloCereda I remember Alf, is it the extra-terrestrial orange monster?

@CarLaTeX Yes!
@CarLaTeX His favourite dish was cats, that's why I remembered him!

@PauloCereda I didn't have a TV until I was in grad school, so no.

@AlanMunn oh
@DavidCarlisle: only Italians and ducks seem to remember this show. :)

@PauloCereda Yes, I remember!

6:51 PM
@CarLaTeX <3

I am very confused... when I do pdflatex and bibtex from the command line I get Warning--I didn't find a database entry for "NSCJML:2017"
but when I load the latex into texstudio it renders just fine without error
what could be going on?

@Lembik As you just said, it's not an error, it's a warning. :)
By error, you mean the compilation being interrupted?
Unless there's an actual error in the .bib file, an undefined reference won't cause pdflatex to halt.
How does the resulting PDF diplay \cite{NSCJML:2017}? Something along the lines of ?? ?

@PauloCereda let me check
it just completes omits it from the bibliography
it really makes no sense that texstudio can render it just fine

@PauloCereda ALF duck

7:08 PM
@samcarter LOL

@Lembik Many TeX editors suppress warnings (or make them hard to find). This, as you've discovered, is not a good thing.
@Lembik Unless I'm misunderstanding the behaviour. In the PDF that TeXStudio creates, does the citation appear in the text, and the reference itself appear in the bibliography? As @PauloCereda says, it should simply generate a ?? (or if you're using biblatex a bold version of the bib file key), and no entry in the bibliography itself. If this is what happened then TeXStudio is simply hiding the warning from you.

@AlanMunn it does appear in the pdf that texstudio makes
which is why it doesn't make sense

@Lembik I see. Is TeXStudio running pdflatex multiple times? Is the warning from bibtex or pdflatex? What happens if you simply run pdflatex another one or two times?

cracked it ... pages = "122<96>-133", in the bib file is what is doing it
errr.. I think
let me see
no it makes no sense at all..still working on it

7:26 PM
@JosephWright colour is fun isn't it:-)
0

If you simplify your example it is easier to see where the space is coming from (and the fact that it is a line of a paragraph, not a space) \documentclass[]{article} \usepackage{xcolor} \usepackage{amsmath} \showoutput \showboxdepth4 \begin{document} 1 \label{eq1} xx \end...

@Lembik either the editor is hiding the warning or it is using a different copy of bibtex search path and finding a different bib file.

ah the path!
let me see
argh!
that's it I think.. BIBINPUTS must be ignored by texstudio
thanks

@Lembik That was my second guess. Although if TeXStudio is getting it right maybe it's your shell path that's not right.

@Lembik it probably sets it to some value in its own config (I know nothing about that editor, it has a fatal editor flaw as far as I am concerned)

@DavidCarlisle It's thinking like that that led to Brexit. :)

@AlanMunn that's not funny (nothing about brexit is funny:(

7:38 PM
@DavidCarlisle I retract the smiley then.

@AlanMunn It would be better if 17million people retracted their vote

thanks very much everyone.. that was driving me crazy
did you see the new OECD report on brexit?

@Lembik if "seen" includes "hear the report on the radio" then yes. Summary as "stating the obvious" ...

7:53 PM
the headline read like a mash report
"Breaking news: Not crashing into Mars seen as good for your health"

@Mico -- how come you deleted your answer on tex.stackexchange.com/q/396673 ? that was exactly the correct answer.

@CarLaTeX it turns out that you don't get sent to jail for going to a pizza shop in the UK after all.
2

@DavidCarlisle They should import that kind of pizzeria in Italy, it would be a huge success LOL

@CarLaTeX they even serve Hawaiian pizza!

8:08 PM
@DavidCarlisle I don't think their customers will mind about the quality of the pizza :)

@CarLaTeX oddly enough most of the customers go there for pizza (and don't land up in court:-)

@DavidCarlisle Because they're English :)

@CarLaTeX better quality of pizza in England do you mean? Yes that seems a reasonable assumption.

@DavidCarlisle Not exactly what I meant :)

8:25 PM
@DavidCarlisle Well yes, of course.

@DavidCarlisle gjer1 and gjer8gj because there is no vowel in gj: gjer-gj, while there is one in gjer-gji

@DavidCarlisle Or what passes for pizza anyway. Domino's pizza is barely pizza completely independent of pineapple. (Also many of us in the US refuse to buy it anyway because the owner is a well-known anti-abortion crusader.)

@AlanMunn yes well i wasn't going to mention those details (well i didn't know the last bit)
@alfred but neither of the patterns that you show are influenced by the vowel i being or not being in gjer-gj, or gjer-gji ?

@AlanMunn -- you've just given me another reason for boycotting domino's. my original reason has been that they don't offer anchovies. (besides, there are much better alternatives locally, ones that even @carla or @egreg would recognize as legitimate.)

8:42 PM
@barbarabeeton Yeah, it's pretty crap pizza anyway so no loss. But doing a bit of googling, it turns out that although this was true, it's no longer true since the owner sold his share in the company in 1998. So you can go back to just boycotting for lack of anchovies and generally crappy pizza. :)

@DavidCarlisle I think gjer1 will separate gjer-gji, and i have 1gji also. I need gjer8gj to stop gjer-gj from hapening
tbh \showhyphens{} gives gjergj and gjergji (no hyphenation) so I should be wrong

@alfred no if the input word is gjergj or is gjergjj then gjer8gj  will match and put 8 after the r and gjer1 will match and put 1 after the r so the 8 will always win and no hyphen will be added, neither pattern looks at the i at all, the same matching would occur for gjergjz

I tried gjer2gj and gjer7 7gji gje7 7rgji, still no hyphenation

oh if you also have gji then yes this will allow a hyphen before gji unless the full string is gjergji so you
@alfred sorry I can not guess what you mean. show a word that should have a hyphen after gjer and one that should not

gjergj gjer-gji

8:54 PM
@alfred and the first does not get a hyphen because no i rather than just because it is within 2 letters of end of word?

The one I need, of course.

the first doesnt get an hyphen because there is only one vowel there

@alfred oh j is a vowel?

e
in the whole word there is only one vowel -e- so there is no hyphenation. each hyphenation segment must have at least one vowel, so you need two vowels to hyphenate a word

@alfred but the i 4 letters later changes that? (I believe you it just seems weird:-) @AlanMunn will be killing himself laughing at my mastery of non english grammar

8:57 PM
@DavidCarlisle The first doesn't get hyphenated because it's a one syllable word, which I assume never get hyphenated in any language.

@alfred aha a rule we like rules:-)
@AlanMunn yes but it helps if you can see the sylables

@DavidCarlisle Well vowels are a pretty good indicator...

gjergji has two syllables gjergji or gje-rgji, both ways can also be hyphanted

@AlanMunn I'm scarred by welsh, I dare not guess which are the vowels

@DavidCarlisle That's entirely understandable. @alfred What language is this, BTW (I haven't been following the discussion.)

8:59 PM
@AlanMunn Albanian

vowels are a e ë i o u y

@AlanMunn .. which has babel setup for fixed language strings but no hyphenation patterns
@alfred thanks

@DavidCarlisle Eek.

gjergj is first name, same as george

@AlanMunn for some strange reason @alfred finds that hyphenating this with US English patterns makes a less than brilliant result:-)

9:02 PM
rules .. albanian doesnt divide some letters

@DavidCarlisle So much for American exceptionalism. :)
@DavidCarlisle But although I might be of help with the linguistic aspects, the dark art of constructing hyphenation patterns is not something I can be of help with.
@alfred Which, in English, despite containing three 'vowels' (orthographically) is also one syllable, and would never hyphenate. @DavidCarlisle :) But assuming Albanian isn't so capricious, vowels can determine syllables (if syllables themselves determine hyphenation.)

I found a system and should be up to 400 patterns soon, although the result isnt bad at 200 already. It's kind of stupid though because there are just three rules
use syllables, dont separate diagram consonants (dh gj nj rr ll sh th zh xh), preferably dont separate diphthongs
diphthongs is not double underwear

Underfull \hbox (badness 10000) in paragraph at lines 22--22
[]  \TU/lmr/m/n/10 gjergj gje-rgji

ok ... how did you do that

@alfred I don't know if it helps but :
\documentclass{article}

%\iffalse % use no patterns for albanian
\iftrue % use new patterns

\makeatletter
\newlanguage\l@albanian
\language\l@albanian
\lefthyphenmin2
\righthyphenmin2
\patterns{
e5rgji
}
\makeatother

\fi
\setlength\textwidth{6cm}
\usepackage[albanian]{babel}
\begin{document}

\showhyphens{ gjergj  gjergji}
\end{document}

9:11 PM
I dont get it, why isnt it separating gjergj like it used to

@alfred by default no hyphenation is allowed, and that pattern says allow a hyphen after e if it is followed by rgji so it only matches the second word

ok, that works, thanks
I also need gje1 for other words though

@alfred so for those you just need 9 patterns d8h, g8j etc, then they will never split
@alfred if you then can match end of syllable to add hyphens you might be able to stop single-syllabel words splitting just by setting \lefthyphenmin and \righthyphenmin or in luatex, the minimum total word length) which would save having to have patterns like e5rgji for every combination of two vowels and a group of non-vowels

yes i did that, also 8d8h8 so they dont end up alone

@alfred but that also prevents a hyphen after dh is that what you need?

9:19 PM
not if there are other letters, is that right ?
dhi would get hyphenated I thought

@alfred no it prevents hyphenation after dh always

so only edhe would get hyphenated
so how do you avoid dh being separated, besides hyphenmin

@alfred for that to be hyphenated you would need a pattern that inserts 9
@alfred d8h not d8h8

how do you use 9 and 0

@alfred same as any other number (but 0 is never used it is the same as not using a digit at all)

9:24 PM
9 goes in the same set as 1 3 5 7 ?

@alfred yes odd numbers enable hyphenation and even numbers stop it, higheset number added at any position wins
@alfred you can also use the letter . to mean a worf boundary so foo. matches foo at the end of a word

ok so i did this : gje1 gjer1 gje2rgj 7rgji and I get gjergj and gje-rgji
does .8d8h8. make sense ?
avoiding .dh- or -dh.

@alfred it means something but it would never be used, tex never adds a hyphen before the first letter of a word or after the last letter so you do not need to stop that

no i mean like dh-emb or mble-dh

@alfred ah hyphenating words that have hyphens is a different issue (not really possible in xetex or pdftex, luatex added some extra features at the last release, but I'd need to remind myself of the details)

9:33 PM
no no i meant if he would divide dhemb or mbledh like this dh-emb or mble-dh
i saw he did it in some cases

@alfred sorry I see what you mean don't you just want to set lefthyphenin and righthyphenmin to 2 or 3, almost all language patterns set 2 or 3 for those values otherwise you need loads of silly patterns to control word ends

if that's the rule as you say then there is no problem, I'll use hyphenmin

If you wanted to do that with patterns you would need separate patterns for .dh and dh. as your suggestion with .dh. would only match the two letter word dh

I might as well add them
I tried, it doesnt give a conflict

@alfred Start using small numbers, 1 for allowing hyphenation, 2 for not allowing it; use 3 for overriding a possible 2, 4 for overriding a possible 3 and so on. If you start with 8 you won't catch exceptions. As far as I know, the main problem with Albanian is that composed words should hyphenate at component boundaries. And this means much more than 400 patterns.

9:45 PM
@egreg in luatex that is probably better addressed by using two sets of patterns a first set just finding word boundaries and a second set hyphenating the compounds and hyphenation callback that does whatever is needed to merge these (doesn't help xetex of course)

that's a different problem. I am working with a linguist on this book, but he's occupied with something else at the moment. He didnt mention the component boundaries, but in principle you are right

@alfred are you using luatex, xetex or pdftex for the actual book?

xetex initially and now lualatex for the patterns

@alfred well the patterns will work fine in xetex as well (but luatex has some extra features) but if you stick to core features your patterns should work in all three systems

I get ! Patterns can be loaded only by INITEX. l.10 \patterns

9:53 PM
@alfred yes that's what the --ini command line switch is for, to allow \patterns but you just want to do that when making xelatex.fmt then xelatex will have those patterns built in just as it does the other languages

in principle if I add --ini to the command my text editor calls I should be fine ?

@alfred no in --ini mode you can only make a format, you can't typeset text.
@alfred which is why (as I tried to explain the other day) for initex or pdftex the system is that you add the language to your language.dat file and then miktex or texlive package manager will use those patterns when rebuilding the format but i don't know the detail of the setup in miktex

I'd like to do that at some point
I found a hyph-it.tex, created a hyph-al.tex with it's content (at some point I might replace that with my own) and two other files connected to them, changed language.dat and rebuild but something is missing

@alfred the split in functionality is because in 1980 a machine w0uld sit for hours (literally) compressing the hyphenation tables so you would not want to do that in a normal document run so the initex version of tex is configured just to do that and dump patterns and macro definitions into a format file and a normal tex run loads the format and processes documents. luatex got rid of the distinction but was written decades later

I was thinking today of the time it must take to recognize the patterns in a 300 page document

10:01 PM
@alfred you probably shouldn't edit language.dat directly as that is under control of miktex so you will just lose the edits on an update. you could ask a question about that or we could ping @UlrikeFischer who knows things

I would stop updating :)

@alfred the trick tex uses is that it doesn't try to hyphenate all words, it does a first pass over a paragraph and then it basically knows the words that could possibly be involved in a line break and only tries to hyphenate those

it makes sense, checking only the words that overflow

@alfred it's a bit trickier than it sounds as it does not work line by line it is doing a least cost calculation over the whole paragraph but it still manages to avoid most words. Luatex actually uses a slightly different algorithm and so can (but usually doesn't) get different hyphenation than xetex or pdftex even using the same patterns as it can find a different set pf feasible breaking words

10:20 PM
the solution to what egreg was saying is extracting patterns from a book, as long as a book can be representative of the whole language
that's what i started doing, reading long words from a book and writing down patterns, now I'm thinking of adding every possible two letter combination containing a vowel, that's 1300 more patterns for a 36 letters alphabet

@alfred of course for a book, you also have the option of using \hyphenation` instead of patterns just have some very basic patterns and then just add every word that needs hyphenation to the \hyphenation exception list over time as you edit the book it builds up a dictionary tuned just to that work

sure, but my friend might ask me doing his other books with latex, so i better have the best hyphenation i can right now. Tbh, the result is already quite good, and i can always use \hyphenation or \- when I need to

@alfred if you get it working you should offer them to the hyphenation group for inclusion in miktex/texlive even if not perfect they must be better than current situation of nothing at all

I'd like to
it's Not done as it should have, but it might be of help

@alfred there are links here:

10:38 PM
I might just upload the file there ?

11:05 PM
what command line can I use ti execute my code ?

11:34 PM
up to 330