note that if you are doing this before starting the table you will need 1 more than \c@LT@tables as incrementing the counter is first thing it does
@nagylzs heiko had an answer on site not so long ago for the simpler case of adding up all the widths (to set the caption) but he wrapped it in a more usable macro I'll see if I can find it
@DavidCarlisle If the name is not defined then mspec is defined as \def\mcspec#1{}, but shouldn't it be \def\mcspec#1{\multicolumn{3}{l}{}} ? Or that does not matter, because I'll have to run latex again anyway.
@nagylzs oops I was changing it at times multicolumn would be better as otherwise the other cells in that row will get set in the wrong columns (so mess up teh column width measuring) I should probably fix that answer:-)
@nagylzs \ifcsname is the "modern" (1990's:-) e-tex way I used the classic TeX \expandafter\ifx\csname test but it comes to the same thing
@DavidCarlisle Thanks. I'm now working on a more general solution that saves all column widths in an array. I have almost no experience in writting latex macros so it will take a while. But I think it will worth the efforts. I have been struggling for this functionality and now it seems possible to do it.
@nagylzs saving all column widths in an "array" is essentially what longtable does (given that tex doesn't have arrays) one for each table. The code I showed was just accessing the widths from that structure (which is what longtable does all the time to make the various "chunks" line up)
@DavidCarlisle I also noticed that this code will set the width to something that was previously calculated. So it will never make the width smaller, only wider. In other words: whenever the content of any cell changes, I have to delete the aux file manually in order to get the desired result. I wonder if small overfull can break this code by widening the column width on each run.
@nagylzs longtable avoids this problem by checking which chunk was used, if you look in the aux files at the LT@... entries each \LT@entry has two arguments the first is a chunk number, the second a length, so when it is aligning chunks in the current chunk it does not use the saved length but uses the length of the current cells, this way if you edit the entry, or even delete a table so that initially it's reading completely the wrong data as the table numbering has changed, it is OK
@DavidCarlisle I see, thank you! I'm not sure how the aux file is generated but I guess the code for longtable writes into this file. What I don't understand is that how defining \LT@entry can "read" these values out of the aux file.
@nagylzs the aux is \input by \begin{document} so at that point all the \LT@iii commands from tables on previous runs are defined so a table doesn't look at the file, the macros it needs are defined (this is also how \ref works : \label writes to the aux that definition is read at begin document so \ref can have it available
@DavidCarlisle Okay, then how \def\LT@entry#1#2 calculates dimen1 ? By the time we define \def\LT@entry#1#2 it was already defined by the aux file right?
@DavidCarlisle Or is it "called" by longtable code somehow, when it tries to set these global LT entries?
@nagylzs LT@entry doesn't calculate anything it is just the "array constructor" the length is its second argument. the way lists work in tex is then you locally define the constructor to do whatever needs doing and execute the list so within longtable LT@entry is defined to make a blank table row with cells of the given widths; in heikos answer it is locally defined to add up the widths; in my answer it was locally defined to add up the first three widths....
@DavidCarlisle I think I understand. It means that if I want to access actual column widths from within the table, they first needs to be saved into an array from the LT@entry constructor. Then I can use that array inside table cells to use their values for calculations freely.
@DavidCarlisle Probably I'll try to save them with arrayjob, and then try to use them.
@nagylzs no but you could easily write an accessor \get{3}{..} if you want (not usually needed and always slower if you do:-) each \LT@i is a list of the form {\foo{}{}\foo{}{}\foo{}{}\foo{}{}} so to iterate over the whole list you don't need for i =1 to n {list(i)} you just execute the list
@DavidCarlisle Yeah, it is clear that I'm not familiar with control structures in TeX. :-( Also not with data structures. I have no idea how to execute a list on a set of incides, for example.
Something like: \sum [3,4,5] \LT@ii
But obviously this is not the way to do it
There are different values for different chunks too. And I would need to specify the chunk of the current cell. Another special value that I don't know how to access :-(
@DavidCarlisle Yes, already had that part figured out. I still feel that I have to save these values somewhere but I also feel I'm wrong in this. :-) In a general solution, I should be able to use multiple invocations on mspec, so I cannot use a global variable called "dimen1". Or can I?
@tohecz well I don't think it's wrong but I think I'd be far more likely to use "maximum value of" (but then I'm English so have no idea at all about English Grammar)
@DavidCarlisle David, your macro has used the first 3 entries. When the table is long, there will be multiple chunks. How can I get the "current chunk of the current cell"? Is there an internal for that?
@DavidCarlisle In other words: \ifnum \count@>0 is not enough. I also need to examine #1 and check if it is in the same chunk. Otherwise it won't work for long tables with multiple chunks.
@nagylzs \c@LT@chunks Otherwise it won't work for long tables with multiple chunks yes it will it records the widest entry in the table (ie the width of the column) and also (for use in some defensive code against table edits) which chunk of the table had that wide entry. But usually it doesn't matter: if you want to span the first 2 columns it doesn't normally matter if the first column is wide because of an entry 10 chunks later
@FlorianMargaine In a system that was really designed for colour that table would have a single colour background panel and the lines would be laid over the top, but colour is hacked into a system not designed that way (I didn't even have a colour screen when writing the color package) and what happens there is that each cell gets a separate panel and the lines go between, but the renderer has heuristics to snap lines to pixel boundaries so specifying adjacent panels is inherently fragile:
@FlorianMargaine just a comment: It looks strange to have a centered column where every line starts with a capital letter. I would suggest either left-aligning (and keeping only the header centered), or not capitalizing
@David: "I didn't even have a colour screen when writing the color package." LOL, a nice one :)
@FlorianMargaine in the pdf the two panels are separated by exactly the width of teh rule, but if the right edge of one is rounded to the pixel boundary on the right and the left one rounds down to the pixel boundary on the left they may touch and squeeze out the rule
I started using TeX in 1987 on SunBSD (later sunos later solaris) using the dviview previewer from the vortex project. That is still one of the best viewers around, you could select individual characters ask what font they were in, and the rendering was as good as the (monochrome) screen could ha...
First of all you should use arrayjobx and not arrayjob (the latter is not really compatible with LaTeX).
Secondly, \let\a\dimexpr will make \a equivalent to \dimexpr and try printing \cws(0)+\cws(1)\relax; then \a will produce an error just like
\dimexpr
\end{document}
You should use \newcomm...
Adding some numbers required 16 lines of code, and the use of an experimental package that should replace \newcommand in the future
I don't get it. :-( Your code also contains 11 lines and an experimental package xparse. Is it really required to use a new package, just to add some numbers?
@egreg The problem must be with my thinking. I'm used to procedural languages and it is making me crazy when I can't just add two numbers with "a+b" with three charaters.
@nagylzs You're not trying to do a+b, in that case: you want to use the array data type that's not directly available in TeX, so one has to build it before using it.
Usually, e.g., when I want to have an if, I do something like this:
if(value) {
... whatever code ...
}
But in LaTeX, its:
\ifnum\count0<100
... whatever LaTeX-code ...
\fi
There are no brackets and so on, for example, which indicate where if-statements begin and where they end. Als...
Just to add my humble POV, difference is in the eye of the beholder. Syntax preference is a matter of taste and ideology; if you grew up accustomed to a certain construct or command pattern, there's a tendency for labeling different representations of the same lagic as wacky or strange. :) Besides, I have the impression Maslow's hammer applies here as well. :) I remember of the arithmetic IF in Fortran and different decimal marks in ALGOL, so your mileage may vary. Personally, I believe this question turns more opinion-based. — Paulo CeredaMay 24 at 11:13
@nagylzs Property lists are the implementation of associative arrays in expl3. A property consists of a key and a value. You can access the value by specifying the key.
@nagylzs Note that what arrayjobx calls arrays are actually property lists (hash variables in Perl)
@nagylzs We're using \newcommand inside a definition, so the argument for the inner macro must be specified by ##1, because #1 denotes the argument to the outer macro. On the site there are questions about this.
I've seen sometimes that people use a double pound sign (##) when defining/using arguments.
What is the difference between the normal argument, #1, and the double sign one, ##1? Are there any restrictions for its use? Can you list the good practices, if any, for this type of arguments.
@JosephWright Yes, and unfortunately it is usually hard for me to understand. My favourite language is Python, and one of the reasons for that is its very clean syntax. I used to use perl before but it was also too cryptic. Here in LaTeX we HAVE TO pollute the local and global namespaces in order to do something useful, and I don't like that
@nagylzs There was a talk at the last TUG meeting on creating a variant of TeX with namespacing, but for the moment we don't have it or really need it: you just have to think in the right way
@nagylzs Remember TeX was designed many years ago and for a specific reason: batch-based typesetting. Knuth didn't expect people to even write 'packages' which applied to more than one document.
@JosephWright I have read TeX tutorials, but when I want to do something that does not already have a package, and when it comes to calculating and programming then it suddenly becomes very hard.
@nagylzs It's a bit like a natural language: if you start with 'this is what I want to do in my native language' and translate it, things are not quite right
@nagylzs I don't know if you are familiar with RPN of Postscript. Some people are way quicker with it but that doesn't mean it is the common way. Python uses the common way but TeX uses a strange state-of-thinking and you need to leave the OOP outside the door. Then computations etc comes slowly to you.
@nagylzs You can do calculations, sure. The 'classical' way is to use assignments and do things one at a time, but nowadays we do have \numexpr, \dimexpr, etc. (e-TeX extensions, added in 1999)
@nagylzs longtable is about 450 lines of code, I think
@nagylzs See for example longtable.dtx versus longtable.sty
@nagylzs If you want a quick insight into TeX at the low level, I'd go with TeX by Topic: texdoc.net/pkg/texbytopic
@nagylzs Expandability and the fact that arguments are passed by TeX 'as is' (not functional style) seem to be the big barriers to people with experience 'elsewhere'
Thank you, very...sexy. Works! I'll have a look into the other suggestions, too. But as usual: who has the time to actually learn Latex? It is the results what one wants, not the way to go there. — Bohlero4 hours ago
Nobody called my solutions/workarounds sexy so far.
I need to use something that looks like a 2 by n matrix as inline math. How can I achieve this without it pushing the lines apart?
Example:
\left(\begin{array}{cccc}
1 & 2 & \dots & N \\
i_1 & i_2 & \dots & i_N
\end{array}\right)
The problem with this code is that it takes as much space as...
I know there are different ways of expressing sizes or dimensions in LaTeX such as points (pt), inches (in) and ex.
As some commands, such as \hspace understand all of them, I would like to have a reference or complete list of possible dimensions or sizes including a description of what they mean.
@tohecz @JosephWright I don't particularly care if it's CW or not. Making it CW won't give me reputation for answer-upvotes, so it was a sort of protection. I was worried that some might think that I posted an extremely trivial self-answer just to get some rep. In reality I couldn't care less about reputation score ...
TeX.SE is pretty friendly though, I'll keep it in mind for the future
It's great that this site managed to stay so friendly despite its size.
Mathematica.SE is much smaller, but I feel that the general mood has been going downhill ... it's hard to tell why. I get annoyed by very basic questions sometimes, so I'm definitely guilty, but I also try to keep in mind what it's like to be a beginner. (In LaTeX I seem to be an eternal beginner which is not a good thing ...)
First of all you should use arrayjobx and not arrayjob (the latter is not really compatible with LaTeX).
Secondly, \let\a\dimexpr will make \a equivalent to \dimexpr and try printing \cws(0)+\cws(1)\relax; then \a will produce an error just like
\dimexpr
\end{document}
You should use \newcomm...
Usually, e.g., when I want to have an if, I do something like this:
if(value) {
... whatever code ...
}
But in LaTeX, its:
\ifnum\count0<100
... whatever LaTeX-code ...
\fi
There are no brackets and so on, for example, which indicate where if-statements begin and where they end. Als...
I've seen sometimes that people use a double pound sign (##) when defining/using arguments.
What is the difference between the normal argument, #1, and the double sign one, ##1? Are there any restrictions for its use? Can you list the good practices, if any, for this type of arguments.
I need to use something that looks like a 2 by n matrix as inline math. How can I achieve this without it pushing the lines apart?
Example:
\left(\begin{array}{cccc}
1 & 2 & \dots & N \\
i_1 & i_2 & \dots & i_N
\end{array}\right)
The problem with this code is that it takes as much space as...
