5:34 AM
Hey!
How can we space correctly expressions with symbols + letters, such as $xy+*z*z*xy*+$? I tried with $x\ y\ +\ *\ z\ *\ z\ *\ x\ y\ *\ +$ but there are some spurious spaces

1 hour later…
7:00 AM
@manooooh What are you actually trying to achieve?

7:48 AM
@mickep hi! What do you mean? I would like to homogeneously space all the characters
Oh, your question helps me to remember that there are a package called soul (I used it) where we can homogeneously space characters!
Anyway do you have native code?

@manooooh Well, I just got curious about in what situation one would like to write math expressions like that.

8:04 AM
@mickep for example, in Boolean expressions, to make trees, to use reverse Polish notation, etc., i.e. in Discrete Mathematics in general
0

Given the boolean expression given in reverse Polish notation $$x\ y\ +\ x\ z\ +\ *\ y\ z\ *\ +$$ recover the tree, write it in usual notation and simplify. The usual notation is \begin{array}{ll} &x\ y\ +\ x\ z\ +\ *\ y\ z\ *\ +\\ \iff&(x+y)\ x\ z\ +\ *\ y\ z\ *\ +\\ \iff&((x+y)+x)\ z\ *\ ...

8:30 AM
@egreg Excuse me very much for last tonight. My connection goes every other day. Right now I'm writing to you from school. I'm so sorry. This morning I checked at home and I'm still without a connection.
@egreg In mt2pro full there is the bold font of the curly script character (first image), but in the green rectangle in the guide there is only the normal character without the bold. What do I have to do to insert it? I also tried David's suggestions, \boldmath, \boldsymbol, etc. but none of them work. Why? For curly script characters I should use bold if possible. Thank you and excuse me for the trouble. Greetings and have a nice weekend.
@Skillmon Good morning and good work.

@Sebastiano and good morning to you!

8:51 AM
@manooooh start by removing all the \ which never really belongs in math mode, then use mathord versions of * and +

The UK patterns are saved!
@JosephWright @DavidCarlisle ^^ less work for you :)

@PauloCereda Quem se importa?

@DavidCarlisle oh /sob

@PauloCereda Yeah, I knew that :)
@PauloCereda If you look at the Debian bug, there was some movement anyway, but I am pleased that the license is being tided up

@PauloCereda Nós preferimos linguagem com marcas aleatórias acima da letra, padrões do Reino Unido não são interessantes.

9:03 AM
@JosephWright oh
@DavidCarlisle oh

@PauloCereda Fancy being chair of UK-TUG? ;)

@JosephWright I nominate Psmith as chair.
3

@DavidCarlisle :)

@JosephWright It wouldn't happen anyway, and the discussions would be lovely. :)
@DavidCarlisle Seconded.

@PauloCereda Someone has to be ...
@PauloCereda @DavidCarlisle's already rebuffed me :(

9:06 AM
@JosephWright I have a name. :)

@JosephWright why?

@JosephWright Brent, perhaps?
I think the Queen should be the honorary chair.

@DavidCarlisle Constitution demans a Chair
@PauloCereda Leaving committee

@JosephWright oh no

@JosephWright I know, but that doesn't mean there has to be a chair. If the committee couldn't be formed it might force some more realistic plans

9:09 AM
@DavidCarlisle Er, yes
@DavidCarlisle I do see your point

@DavidCarlisle Psmith now seems like a good idea, specially with 8ball...

@PauloCereda he needs to agree to be chair, but if you ask enough times he probably would

@DavidCarlisle ooh I like the odds
As the saying goes, "There are three kinds of lies: lies, damned lies, and statistics." We know that's true because statisticians themselves just said so. From a report: A stunning report published in the Annals of Internal Medicine concludes that researchers often ask statisticians to make "inappropriate requests." And by "inappropriate," the authors aren't referring to accidental requests for incorrect statistical analyses; instead, they're referring to requests for unscrupulous data manipulation or even fraud. The authors surveyed 522 consulting biostatisticians and received sufficient r

9:35 AM
@DavidCarlisle I'll prod the committee on this: yoiu do have a point
@DavidCarlisle Perhaps you might send me something to raise at AGM, if you are not coming yourself

@JosephWright no I'll probably be there

@DavidCarlisle Cool
@PauloCereda Will you have a talk?

@JosephWright Sign me in.

10:08 AM
@JosephWright Phil :(

@DavidCarlisle Yeah
@DavidCarlisle Misses the point: the Knuth 'license' doesn't stand up to legal scrutiny
@DavidCarlisle Not his call anyway
@DavidCarlisle, @PauloCereda I've poked the committee about the Chair role and wider context

@JosephWright yes debian only accept it as they believe it's invalid....

@JosephWright let slip the dogs of war?

@DavidCarlisle I know
@DavidCarlisle Point is, the other people who use hyphenation patterns do accept MIT

@JosephWright yes

10:11 AM
@DavidCarlisle Also, hyphenation patterns are not inviolate for ever, otherwise we'd be really stuck with names ...
@DavidCarlisle Anyway, Dominic is sensible, so issue goes away

@JosephWright I hope so, except he explicitly said to consult Phil...

@DavidCarlisle Ah, there is that
In other news: hits on my blog are rising since I installed some SEO stuff
I need to work out how to pay for hosting: it's gone up :(

@JosephWright woo
@JosephWright oh no

@Sebastiano Sorry, I don't have mt2pro
@manooooh Make all symbols belong to the punctuation class.

10:23 AM
@JosephWright you could earn advertising revenue by covering the site in full width download banners for unrelated software

@DavidCarlisle Yes ... or I could just put my hand in my pocket
I like plan b

@JosephWright get uk tug to use some of the money it doesn't know what to do with to support dissemination of tex information?

@DavidCarlisle Yeah, I can really see that working

@JosephWright it seems more in line with the constitutions stated aims than leaving the money in a bank
@egreg I suspect @manooooh wanted a mathjax answer to use on the math site here..

@DavidCarlisle Some people will see it as personally-beneficial to me

10:28 AM
@manooooh Here's an implementation

@DavidCarlisle I did consider it: I might still do

\documentclass{article}
\usepackage{amsmath}
\usepackage{xparse}

\ExplSyntaxOn
\NewDocumentCommand{\RPN}{m}
{
\text{$\manooh_rpn:n { #1 }$}
}
\cs_new_protected:Nn \manooh_rpn:n
{
\tl_map_inline:nn { #1 } { \mathpunct{##1} }
}
\ExplSyntaxOff

\begin{document}

\RPN{xy+xz+*yz*+}

\end{document}
@DavidCarlisle You could implement expl3 in JavaScript

@DavidCarlisle I do get people suggesting that (people who sell ads, of course)

@JosephWright is that who you use now, or who you might switch to?

@DavidCarlisle Who I use now: they are pretty good
@DavidCarlisle Better I think that the previous people

10:31 AM
github pages are free:-)

@DavidCarlisle I know, and if I was starting from scratch I'd likely go that way
@DavidCarlisle Migration was tricky last time I looked

@DavidCarlisle ducky bank
@JosephWright you could stream games. :)

@JosephWright probably yes, especially for comments etc. which gh-pages doesn't really do so you'd have to use disqus or some such

@DavidCarlisle I was going to suggest Disqus.

2 hours later…
12:16 PM
@JosephWright What SEO stuff did you install? On the other hand, it looks like a normal increase, during the two years

1:15 PM
OK. There are some differences. The differences are: article.cls 2014/09/29 v1.4h, size10.clo 2014/09/29 v1.4h, xparse.sty 2018-09-24, expl3.sty 2018-09-24, expl3-code.tex 2018-09-24, l3pdfmode.def 2018-09-24, inputenc.sty 2018/04/06 v1.3bDavid Purton 10 mins ago
@DavidPurton ^

@DavidCarlisle, I think it is just page size
@DavidCarlisle What default page size do you use? Debian (in it's wisdom) allows me to set the default page size to A4
@DavidCarlisle But I often forget that when answering questions for others who might use the standard default of Letter. In the code under question, it matters because the column is narrower

@DavidPurton that should never make any difference to tex's layout it is just the default external page size in the pdf, \textwidth, textheight and friends always have the same values (based on US letter)

@DavidCarlisle, in fact I have a geometry.cfg containing: ExecuteOptions{a4paper}

@DavidPurton ah well there you go, that will be it:-) (I'd say it's a bad idea to have that config at all, just makes your documents non-portable)

@DavidCarlisle, I'll update my answer to the question with a specific paper size

1:26 PM
@DavidPurton thanks, can you answer todays question about the difference as well.

@DavidCarlisle will do. Believe it or not, no one else in my circles uses LaTeX for me to need to create portable files ;)

@DavidPurton unless you post tex documents to half a million stackexchange users:-)

@DavidCarlisle true. I'll think about changing it ;)

@DavidPurton I have write access to geometry these days, I could remove it... (but won't)

Today I saw a man driving a car with UN IX on its license plate and two ducks in the front, @PauloCereda
4

1:37 PM
@DavidCarlisle It wouldn't help. I think Debian sets the default classes default paper size to A4 as well when you set your system wide preferred paper size.
@DavidCarlisle In the end, the only sensible solution is to make A4 default for all the LaTeX classes ;)

@DavidPurton no the default classes have no mechanism to do that unless debian have changed the sources, which seems unlikely. This is by design, portability was considered more important than the minor inconvenience of having to use [a4paper]
@DavidPurton and for that reason geometry .cfg is a bad idea and while I can't remove it I probably will add a warning to the documentation suggesting that it not be used.

@DavidCarlisle No, this is not right. On Debian a simple article based document with no packages loaded produces an A4 document for me.

@DavidPurton that would be a bug (and I doubt it's true:-) The pdf media size will be A4 but the layout will be designed for US letter unless you add [a4paper]
\documentclass{article}
%\documentclass[a4paper]{article}

\begin{document}

\showthe\textheight
\end{document}
@DavidPurton you should get 550pt or 598pt depending on whether [a4paper] is used.

1:55 PM
@DavidCarlisle OK, you win. It is just the PDF media size that's set to A4. Hmmm. That's quite confusing behaviour I admit. I simply assumed that the layout was A4 if the media size was A4.

@DavidPurton I was there when that code was written:-)

@DavidPurton the design predates widespread on screen viewing. dvi has no concept of "media size" there is just the text block size and the media size is whatever size paper is in your printer.

2:15 PM
@DavidCarlisle I can see how that sort of makes sense... Anyway, I updated my original answer and answered the follow up question.

2:33 PM
@egreg up to your old tricks again I see.

3:07 PM
@DavidCarlisle He spotted a space?

@HaraldHanche-Olsen no took my answer added a bit of superfluous expl3 showiness and ran off with the tick:-)

5

3:39 PM
@UlrikeFischer for Chair!
everyone got their tugboat?

@StefanKottwitz Just YOAST: to be honest it's more a gentle rise. But it wsa higher a couple of years before

@JosephWright A provider charging per page hit (instead of volume, performance etc.) sounds a bit like a ripoff to me.
@JosephWright At any time you can put the site on the server that's supported by DANTE. A dozen Wordpress sites run there already, one more doesn't mean more operational work or costs.
@JosephWright No provider fee, no advertisement needed, not another user group to pay
@JosephWright Daily backup and backups pushed to a site of your own if you like

@StefanKottwitz How will UKTUG ever manage to spend its money if people keep offering things for free? :-)

@DavidCarlisle Spend the money on professionals doing work for TeX who are not TeX friends doing it for free.

@StefanKottwitz but they don't do anything so it's hard to know which professionals they could employ

3:53 PM
@DavidCarlisle Just a sample where a professional book cover designer could have helped I guess, instead of doing it by a TeX friend for free:
I guess I better delete that rant :-)

@StefanKottwitz that was tug though not uktug:-)

@DavidCarlisle Yes, just a sample. Really spend money on key things or promotional things but internal or routine stuff can be done for free saving money for key things

@StefanKottwitz but they don't do anything so there are no routine things and no things to promote professionally. Just an AGM once a year and a chance to buy everyone a pizza.

@StefanKottwitz Who said they were charging per hit?
@StefanKottwitz Can't do that: being an employet = complex tax rules
@StefanKottwitz It's always been hard to find things that can be paid for that don't involve, one way or another, paying the people asking for the money with cash that goes into their pockets (salary, consultancy fee, ...)

@DavidCarlisle Like the TUG or DANTE sponsored projects, developing some good stuff that would otherwise not exist, or really needed things such as a better LyX UI or other stuff we are not good at :-)
@JosephWright you posted a hit counter (view per day) and I saw hit limitations in the site plans, so I assumed
@JosephWright I sometimes hired people via fyverr.com to do things better than I can and to save time

4:08 PM
@StefanKottwitz Ah, right

That site sounds cheap and jobs sound often too cheap but anyway one can find professionals and pay easily through the site and do it fair.

@StefanKottwitz They do have a 'soft' limit on executions: I got prodded toward my current plan because of the total traffic
@StefanKottwitz Interesting ...

@JosephWright There one can find people doing design, translation, programming, whatever. The rest is up to negotiation.

@StefanKottwitz Still (probably) wouldn't be allowed by the UK-TUG committee: it involves paying people, not for services, goods, etc.

@JosephWright Right, it's for peoples' work

4:15 PM
@JosephWright any thoughts on this diff I get from @UlrikeFischer's tlg, don't we remove white space from the start of the line usually?
*** ./build/test/hycolor-test1.xetex.tlg        2018-11-03 15:07:10.865520500 +0
000
--- ./build/test/hycolor-test1.xetex.log        2018-11-03 15:07:13.054633600 +0
000
***************
*** 26,32 ****
Package qstest Info: Passed: HyperrefColor on input line ....
Package qstest Info: Passed: X0134 with xcolor on input line ....
Package qstest Info: Passed: FieldColor on input line ....
! (hycolor-test1.aux)
LaTeX Font Info:    Checking defaults for OML/cmm/m/it on input line ....
LaTeX Font Info:    ... okay on input line ....

@DavidCarlisle We used to: I think I stopped that as it led to more diffs

@JosephWright ah the actual problem is the long line previous, the raw log is:
(/usr/local/texlive/2018/texmf-dist/tex/context/base/mkii/supp-pdf.mkii
\scratchcounter=\count89
\scratchdimen=\dimen103
\scratchbox=\box26
\nofMPsegments=\count90
\nofMParguments=\count91
\everyMPshowfont=\toks17
\MPscratchCnt=\count92
\MPscratchDim=\dimen104
\MPnumerator=\count93
\makeMPintoPDFobject=\count94
\everyMPtoPDFconversion=\toks18
) (./hycolor-test1.aux)

@DavidCarlisle I could perhaps add a stop before the aux is read in such cases.

@UlrikeFischer yes I was looking at that, in some cases we could do that, in others the structure of Heiko's tests make it a bit tricky to insert the l3build directives, but actually there are not so many failures. In a couple it might work simply to load tikz before the \START so that there are less package loading lines in the tlg with different wrapping
$find . -name \*diff ./build/test/hycolor-test1.luatex.diff ./build/test/hycolor-test1.pdftex.diff ./build/test/hycolor-test1.xetex.diff ./build/test-config-noxetex/embedfile-test3.luatex.pdf.diff ./build/test-config-noxetex/embedfile-test3.pdftex.pdf.diff ./build/test-config-noxetex/pdfcolparallel-test1.luatex.pdf.diff ./build/test-config-noxetex/pdfcolparallel-test1.pdftex.pdf.diff ./build/test-config-pdftex/pagegrid-test1.pdftex.diff ./build/test-config-pdftex/selinput-test1.pdftex.diff ./build/test-config-pdftex/selinput-test2.pdftex.diff @UlrikeFischer the hycolor ones are all spaces before ( the selinput ones are because I have a newer utf8enc.dfu it's tempting to add some \OMIT\TIMO but as the test is testing encoding switches it's not so clear what can be omitted) and the pdf ones seem to have different timezone information, are we missing some of the EPOCH setting flags? @DavidCarlisle Yes, I guess we will have to insert/move the start/stop a bit. Btw: I have no idea what to do with hologo-test-list (I moved it to "testfiles-unclear" for now). It creates 64 pages of pdf, tests about 10 fonts and also the output in bookmarks. \showoutput gives a log of around 2MB, and pdf comparision doesn't sound sensible either ;-(. @DavidCarlisle the pagegrid-test still fails for me (when I switch the computer) because of an absolute path in the log. 4:25 PM @UlrikeFischer I knew there was a reason I'd avoided looking at these tests:-) @UlrikeFischer it fails for me but wit diffs like: Package xcolor Info: Model wave' substituted by hsb' on input line .... ) ! (../pgfcore.code.texPackage: pgfcore ....-..-.. v....1a (rcs-revision 1.7) (../pgfmath.code.tex (../pgfmathcalc.code.tex (../pgfmathutil.code.tex) ! (../pgfmathparser.code.tex\pgfmath@dimen=\dimen114 \pgfmath@count=\count... \pgfmath@box=\box... \pgfmath@toks=\toks... *************** *** 187,193 **** ) (../pgfcoreimage.code.tex File: pgfcoreimage.code.tex ....-..-.. (rcs-revision 1.18) ! (../pgfcoreexternal.code.texFile: pgfcoreexternal.code.tex ....-..-.. (rcs-re @DavidCarlisle ;-). But imho 95% would be a good quote, better than the 0% now. Also there is some interesting stuff in it. E.g. all the qstest stuff with "Expect"-value testing. @UlrikeFischer yes. thanks for all the work you've done on this. For the pagegrid one would it helo to load tikz before the start to get rid of all the pgf loading from the comparison @DavidCarlisle looks like one should avoid to get too many package loading in the tlg ... @UlrikeFischer for the hologo list it has visual posibilities to test that things look good but if we ignore that it tests the logos work in bookmarks etc so I'd just set it up with \START\hologoList\END and so an expected more or less empty tlg file, and it just checks for no-errors. 4:46 PM @egreg wow it worked, thanks!! Three questions: 1) is it possible to put it in display mode? I tried with \tl_map_display:nn { #1 } { \mathpunct{##1} } but gives error; 2) can we change the spacing to one with more spacing? 3) Do you want me to post a question? ;) @DavidCarlisle it is not necessary, I need it for a document, although it is always better to be nice wherever it is published ;) @egreg well for 2) I wrote \tl_map_inline:nn { #1 } {\quad\mathpunct{##1} } and it works. Is it well coded? @manooooh the "inline" there does not refer to inline/display math but to inlining the function. The same code would work in display math , mathpuct adds \thinmuskip so you can set \thinmuskip to whatever you need @manooooh the coding is OK but isn't the output rather weird with a \quad space, it means that your operators will be nearer following data than the preceding data to which they apply? a b + c + @DavidCarlisle where can we write \thinmuskip? Consider this MWE: \documentclass{article} \usepackage{amsmath} \usepackage{xparse} \ExplSyntaxOn \NewDocumentCommand{\RPN}{m} { \text{$\manooh_rpn:n { #1 }$} } \cs_new_protected:Nn \manooh_rpn:n { \tl_map_inline:nn { #1 } {\quad\mathpunct{##1} } } \ExplSyntaxOff \begin{document} Test \RPN{xy+xz+*yz*+} hello \end{document} @DavidCarlisle @egreg ^^^^^^^^ I want something like this @DavidCarlisle yes, you are right I like this: \tl_map_inline:nn { #1 } {\;\mathpunct{##1}\;} @manooooh I wouldn't use math mode at all, as you just want equal spaced characters whether they be x or + so the math spacing isn't really helping. @DavidCarlisle @JosephWright From LaTeX team point of view, would it better if the HarfBuzz font loader was compatible with XeTeX font syntax or with luaotfload’s (keeping in mind that not everything luaotfload supports makes sense for HarfBuzz)? @DavidCarlisle as I said, this is for a document, so when I show it I would like to display it in math mode 4:59 PM @manooooh remove the mathpunct so you get equal space, but don't you want the same for x an dy? @KhaledHosny in practice fontspec hides the differences and almost no one uses the primitive syntax in either form, so probably whatever works best for you.... @DavidCarlisle with \tl_map_inline:nn { #1 } {##1} there is no same space between letters and symbols @DavidCarlisle Good. @manooooh oh sorry yes. @DavidCarlisle I added an \RequirePackage{tikz} before the test. @UlrikeFischer thanks. @manooooh \;{##1}\; 5:03 PM @manooooh If you want more space, change the value of \thinmuskip. A general question might be interesting. @egreg that was my first reply, but actually wants symmetric spacing... @egreg oh skip that, my first reply was OK:-) Hmmm... some Boolean expressions have logical symbols, so if I write Test \RPN{xy+xz+*yz*+\implies} hello there is more space between + and \implies than the other symbols. I know \implies has extra space... but I wonder if we can forgot that extra space @manooooh because it is applied to all the symbols the first suggestion of using thiunmuskip is probably better @DavidCarlisle I do not know how to change the value :(. Let me post a question and you answer it I have added a comment (the last one) under the answer tex.stackexchange.com/a/35696/63097, which summarizes all other comments, except the penultimate. I flagged some of these comments (which my comment summarizes), but I ran out of flags for today. So, if you have the permissions, you can delete them and just leave the two last comments, I think. 5:06 PM @DavidCarlisle Hmm, a simple LaTeX document that does not use any packages seems to load luaotfload when processed by LuaTeX. @DavidCarlisle I will look at the other differences later. But if you want to change something youself you can sent a PR, or I can add you to the repo. @KhaledHosny Yes, but we deal with that, and it's just for Latin Modern @manooooh \Longrightarrow not \implies` @DavidCarlisle but you guys told me that \implies is better than others... @manooooh \implies is defined as \;\Longrightarrow\; and intended for traditional infix layout so exactly what you don't want here. 5:09 PM @JosephWright So the font syntax I pick will have effect on core LaTeX, right? (assuming it would want to use my font loader, of course). @DavidCarlisle ok... @KhaledHosny Not really: provided we can pick up that it's 'BuzzTeX', we can adjust latex.ltx, and Will can fix fontspec @manooooh Try this \documentclass{article} \usepackage{amsmath} \usepackage{xparse} \ExplSyntaxOn \NewDocumentCommand{\RPN}{m} { \text { \manooh_rpn_if_display:T { \thinmuskip=12mu }$\manooh_rpn:n { #1 }$} } \cs_new_protected:Nn \manooh_rpn:n { \tl_map_inline:nn { #1 } { \mathpunct{##1} } } \prg_new_conditional:Nnn \manooh_rpn_if_display: { p,T,F,TF } { \use:c{if@display} \prg_return_true: \else: \prg_return_false: \fi: } \ExplSyntaxOff \begin{document} \RPN{xy+xz+*yz*+} $\RPN{xy+xz+*yz*+} @JosephWright If things went well, there will be no BuzzTeX, just regular LuaTeX. But I guess you can try require("luaharfbuzz") and work based on the result. @DavidCarlisle I tried that but it gives me errors 5:13 PM @KhaledHosny are you planning to avoid a separate binary compilation, or do you mean hans/Luigi may accept some of your changes? @DavidCarlisle I haven’t talked with Luigi directly yet, but I’m told he is supportive of the idea, so a single binary is a possibility. @egreg yess!! I do not why in inline mode you delete the space but I am sure that I never write this kind of expressions in that mode. Thank you! @KhaledHosny that would be excellent if it happens @manooooh Between two consecutive punctuation symbols in math mode, TeX inserts a thin space (in the amount of \thinmuskip). @manooooh If you want a bit more space in inline formulas, change to \manooh_rpn_if_display:TF { \thinmuskip=12mu } { \thinmuskip=6mu } @egreg now my idea is to write that expressions in another notation. For that, I add (...) inside \RPN{xy+xz+*yz*+}. For example: \[\begin{array}{ll} &\RPN{xy+xz+*yz*+}\\ \iff&\RPN{(x+y)xz+*yz*+}\\ \iff&\RPN{((x+y)+x)z*yz*+}\\ \end{array}$ @egreg ^^^^ as you can see, each time the text gets longer. My question is, is it possible to make a conditional that evaluates that if there are parentheses then the spacing between them is smaller? 5:20 PM @manooooh That's more complex. @manooooh cant you use some other symbols for the infix versions? it is rather confusing to have + being used with two incompatible syntaxes in the same line @egreg yeah I supposed... Why did not you wait for me to ask a question on the site? Hahaha @manooooh your second step seems wrong, the + applies to x z @DavidCarlisle what? Are you saying that the expression is not a valid expression to write it in other notations? Anyone interested? 5:24 PM @DavidCarlisle you do not in what notation I want to write it... It is in reverse Polish notation, where the subtrees are first traversed in their order (from left to right) and the root at the end @manooooh in the second line the first + is infix, applying to x and y and the second + is postfix applying to x and z, which is confusing to say the least. (also then in the third line you make the secon + apply to {x+y) and z If you use VSC, then maybe you're interested @manooooh I know what reverse polish is, but you are transforming it to normal inline syntax using the same symbols. I've not tried it yet, though @DavidCarlisle yes, in math.SE a user told me that, it should be (x+y)(x+z)*yz*+ and so on 5:27 PM @manooooh that's clear but you should use (say) bold operators for the reverse polish and normal for normal, it makes no sense to convert it step by step using the same symbols for the converted and non converted parts. @DavidCarlisle it should be: @manooooh simpler to think of RP as a stack than a tree traversal. @DavidCarlisle actually, the book that I'm guiding myself in does the same way, just comment on each step. I did it in one place with iff and, obviously, no commenting @manooooh I still think it would be clearer if you distinguished the symbols, but whatever. @DavidCarlisle what symbols? Every step of the expression should contain bold symbols and normal symbols? 5:34 PM @manooooh look at your second line the first + has a completly different meaning to the second and third + . From the context it's possible to see what you mean but it is harder than it need be. @DavidCarlisle so the first step would not have any bold symbol. The second step would have (x+y) bolded, right? If so, when I write (...) \iff&\RPN{\bm{(x+y)}(x+z)*yz*+}\\ (...) (bm package loaded) it gives me 100 errors (that's because I think RPN does not support that kind of command) :( @manooooh blame @egreg for not making RPN bm compatible but actually that's the whole point, after the first line you are not using RPN so why push the infix parts through the RPN macro? anyway we have to go out to celebrate torturing prisoners 2 @manooooh For a full fledged function, some specifications are needed. @DavidCarlisle tic toc boom? 5:41 PM @egreg oww <3 do not worry! It works perfect, @DavidCarlisle opens my mind with other stuffs......... $\begin{array}{ll} &\RPN{xy+xz+*yz*+}\\ \iff&\bm{(x+y)}\RPN{xz+*yz*+}\\ \iff&\bm{(x+y)(x+z)}\RPN{*yz*+}\\ \iff&\bm{(x+y)*(x+z)}\RPN{yz*+}\\ \iff&\bm{(x+y)*(x+z)(y*z)}\RPN{+}\\ \iff&\bm{(x+y)*(x+z)+(y*z)} \end{array}$ @DavidCarlisle ^^^^^^ happy? 1 hour later… 6:47 PM @DavidCarlisle I had a look: my expl3 code is ctablestack, essentially 7:14 PM @DavidCarlisle I see that Mojca is, as expected, a voice of reason (the key being that stability is really a social, not a legal, issue) 7:47 PM Does anybody know how to plot 1/x using declare function with PGFPlots? I tried declare function={f(\x)=ifthenelse(\x==0,0,1/\x);} but an ugly line connecting the two parts appears :( @manooooh well I'd just have changed the operators as x y and z are the same, that's rather the point of the transform @DavidCarlisle ok, () include too? @manooooh possibly not, It wasn't a fully worked out plan, it was just an "initial thought" on seeing your output:-) @DavidCarlisle lol :) 8:29 PM @DavidCarlisle for embedfile-test3 it could be the mode date of the embedded file that leads to problem (and probably @JosephWright will have to do something here), but I see nothing in pdfcolparallel-test1.tpf. Can you should me the diff? 8:42 PM *************** *** 1,8 **** %PDF-1.5 %�� 1 0 obj ! <</Type/EmbeddedFile/Subtype/plain#2Ftext/Params<</ModDate(D:20180328124306+02 '00')/Size 66033/CheckSum<AAE437B88FF8AAE581A840818AAE92E2>>> ! /Length 66033 >> stream % \iffalse meta-comment % --- 1,8 ---- %PDF-1.5 %�� 1 0 obj ! <</Type/EmbeddedFile/Subtype/plain#2Ftext/Params<</ModDate(D:20181103123605Z)/ Size 63842/CheckSum<C14E6763E6214561DF5FE5E45A88E19A>>> ! /Length 63842 >> stream % \iffalse meta-comment % @UlrikeFischer ^^ @DavidCarlisle yes, that's the moddate of the embedded file in the embed test. But what is the problem with pdfcolparallel-test1.tpf? *** ./build/test-config-noxetex/pdfcolparallel-test1.tpf 2018-11-03 12:37:02.450316700 +0000 --- ./build/test-config-noxetex/pdfcolparallel-test1.pdftex.pdf 2018-11-03 12:37:05.434971500 +0000 *************** *** 2,8 **** %ÐÔÅØ 3 0 obj << ! /Length 3368 >> stream 0 g 0 G --- 2,8 ---- %ÐÔÅØ 3 0 obj << ! /Length 3385 >> stream 0 g 0 G *************** *** 10,34 **** 0 g 0 G 0 1 0 rg 0 1 0 RG BT ! /F8 9.9626 Tf 148.712 657.235 Td [(Green)]TJ 0 g 0 G 0 1 0 rg 0 1 0 RG @UlrikeFischer ^^ @DavidCarlisle Hm. your pdftex seems to use one decimal more than mine. @UlrikeFischer that's a stock pdftex from tl2018, It is funny when you realized that Arabic script dropped hyphenation-like feature as it matured, but LGC scripts are still holding onto it. Had they dropped hyphenation, I’d have had a more productive and less grumpy week. 8:57 PM @DavidCarlisle mine too. I just rechecked. I hadn't used miktex or so. @DavidCarlisle Cygwin build? @UlrikeFischer, @DavidCarlisle System-dependent variation ... @JosephWright yes @JosephWright makes pdf testing a bit fragile though... @DavidCarlisle Hmm @DavidCarlisle Might have to do some rounding in the PDF then could ask on texlive list I suppose... @DavidCarlisle Sounds good 9:09 PM @JosephWright rounding will get you so far but then all the byte offsets are different so you'd have to lose the xref table completely @DavidCarlisle Ah right, of course @DavidCarlisle We've not seen issues thus-far for other PDFs: I wonder if this is a common issue @DavidCarlisle, @PauloCereda No biters on the committee list thus-far @JosephWright Psmith it is then. @DavidCarlisle Sounds good to me @DavidCarlisle See email @JosephWright the patterns one? @DavidCarlisle Yes 9:26 PM @JosephWright @DavidCarlisle do you have an idea what to do with the /ModDate? <</Type/EmbeddedFile/Subtype/plain#2Ftext/Params<</ModDate(D:20181103123605Z)/‌​. Imho it wouldn't work to simply remove it with l3build. the length and the xref will be wrong. Currently the only idea I have something like \def\pdffilemoddate#1{} which looks a bit drastic. @UlrikeFischer It gets fixed by the reproducible build stuff @JosephWright not for the embedded file, that's the problem, it is written by \pdffilemoddate and this not affected by SOURCE_DATE_EPOCH. Well I guess, if I rewrite embedfile I will have to add a switch, that suppress the moddate if needed. @UlrikeFischer The external file should be fixed, though, or generated using another reproducible build step @UlrikeFischer I was going to say same, I need to check when it got built shouldn't the environment already have the fixed epoch settings? @JosephWright the external file itself is not the problem (it is a local copy of embedfile.dtx). But embedfile writes a dictionary to the pdf with \pdffilemoddate{embedfile.dtx}. Simply try \documentclass{article} \begin{document} \pdffilemoddate {article.cls} \end{document} My output is D:20180529230632+02’00’, but yours is probably different ... 9:39 PM @UlrikeFischer That would be a flaw in the reproducible build stuff: one to ask about on the TL list and get the engine people to fix it @JosephWright yes, it should be possible to force a time zone here. @JosephWright in luatex it should be rather straightforward to solve the problem: it is a function in oberdiek.pdftexcmds.lua and one only need a test for SOURCE_DATE_EPOCH. <-- @DavidCarlisle. @UlrikeFischer Ah, right @UlrikeFischer But needs fixing in pdfTeX @JosephWright not to mention xetex:-) @JosephWright I will sent a message to texlive. @UlrikeFischer @JosephWright it would presumably work to touch the file to give it a known moddate before the test. 9:53 PM @UlrikeFischer Great @DavidCarlisle That's true ... on *nix @DavidCarlisle the problem is not the date but my time zone, the +02’00’ at the end. @UlrikeFischer Probably got overlooked @JosephWright but in general I'm not sure a reproducible build epoch setting is expected to fix this, you can "set the clock" to a known date but if you fake all file dates to teh same date make files and other things tracking if generated files are newer than sources are going to be broken @UlrikeFischer you should live in Z time zone. @DavidCarlisle Could be: I guess the general case is that they use a fixed timezone @UlrikeFischer but on my diff it isn't just the timezone, the whole date is different? 10:05 PM @DavidCarlisle That must be due to the git handling. Locally the date is still fine (I copied it in the testfiles-noxetex/support folder), but what date does it now have on github? @UlrikeFischer git is rather rubbish at preserving filedates... @JosephWright and powershell apparently @DavidCarlisle Ah, yes, of course @DavidCarlisle I mean, I have touch in the Command Prompt via GOW, but that's not general @JosephWright I have lots of things in cygwin and in the lnux subsystem, but i assumed powershell is more in the spirit of things @DavidCarlisle Yes, we can I guess assume it @JosephWright or it seems lua lfs has lfs.touch 10:15 PM @DavidCarlisle great ;-). @UlrikeFischer I guess these dates don't match yours$ ls -l testfiles-noxetex/embedfile-test*
-rw-r--r-- 1 davidc Domain Users  1379 Nov  2 00:56 testfiles-noxetex/embedfile-test1.luatex.tlg
-rw-r--r-- 1 davidc Domain Users  2248 Nov  2 00:56 testfiles-noxetex/embedfile-test1.lvt
-rw-r--r-- 1 davidc Domain Users  1584 Nov  2 00:56 testfiles-noxetex/embedfile-test1.tlg
-rw-r--r-- 1 davidc Domain Users 70788 Nov  2 00:56 testfiles-noxetex/embedfile-test3.luatex.tpf
-rw-r--r-- 1 davidc Domain Users   549 Nov  2 00:56 testfiles-noxetex/embedfile-test3.pvt

@DavidCarlisle no. my test1 is older (from Nov 1), and the test3 is now newer as I played around with it in the last hour ;-).

41

YES, metastore or git-cache-meta can store such (meta-)information! Git by itself, without third party tools, can't. Metastore or git-cache-meta can store any file metadata for a file. That is by design, as metastore or git-cache-meta are intended for that very purpose, as well as supporting ba...

@DavidCarlisle @JosephWright while changing the time of the embedded file with touch may work here, I don't think that it is the right solution in the long run e.g. when files are embedded that are first created with filecontents (and I assume that with tagging I will have such cases). Imho it is better to change the output of \pdffilemoddate/\pdf@filemoddate in such tests.

@UlrikeFischer Agreed

10:30 PM
@UlrikeFischer I'm not sure, it would be equivalent to doing \def\pdfmodfiledate#1{00000} but would mean you would get reproducible test results but wouldn't be testing that the command returned the modification date of a file (@JosephWright)

@DavidCarlisle Yeah ...

That is rather different from the other epoch settings which more or less do a normal run as if the system clock was set to a specified time.

@DavidCarlisle yes a general redefinition looks drastic, probably one will have to control the places where such modifications are needed - so I should rewrite embedfile to inject a boolean ;-).
@DavidCarlisle also if git changes all file dates then reproducible testing in this area could be a bit difficult. Which files can we all access that have a known date which isn't changed by git, or tlmgr, or ... ? Probably we would need something like touch here too.

@DavidCarlisle Reading Travis-CI deployment docs ...

10:49 PM
@JosephWright for saving test diffs?

@DavidCarlisle For Chris's request

@JosephWright ah. I did something similar for openmath pushing some files built by travis in one repo to a separate github pages site, it didn't use the specific "deployment" settings just did a git commit "by hand" in the script running on travis but basically needed the same authorisation setup.

@DavidCarlisle Yes, in the end it could go 'wherever'
@DavidCarlisle As we only want docs building in some cases (certainly after tests have run, but not claiming a test failure for doc issues), probably I can use a custom script

@JosephWright or could we do this so the dtx files appeared in overleaf (which I have never used:-) overleaf.com/learn/how-to/…

11:11 PM
@DavidCarlisle I see how GitHub releases work now :)
@DavidCarlisle dev.topheman.com/continuous-deployment-with-travis-ci looks good: can use say Surge, do a post-check doc and send all of the docs in one big zip

11:51 PM
@UlrikeFischer on my system at least this makes a test file from 1970 (the 0 could be replaced by the number from the epoch setting or anything else)
file = io.open("test.txt", "w")
file:write("Hello World")
file:close()
lfs.touch("test.txt",0)