TeX typesetting

I want to produce a PDF that looks good on the screen in color. Of course if I do that well with color backgrounds and all, the same document will look lousy when printed on a monochrome laser printer. E.g. consider a text box with color background. The background will go through a dithering algorithm which often enshitifies the text layer on top of that. Likewise on mono e-readers.

In principle, the doc needs two different representations. One for color and one for mono. As rich as the PDF standard is, I don’t think I have ever seen a PDF with multiple modes. So LaTeX aside, does the PDF standard even support this?

I can think of a hack using PDF layers which is supported by the ocgx2 LaTeX package. Color backgrounds could be isolated to a switchable layer. This is not great though because the end user needs to be aware of the layer and must take a manual action to turn off the background layer before printing as black and white. And still, non-black foreground text will print as gray unless foreground text is in a layer too (yikes).

Am I S.O.L?

URLs can be long and ugly as fuck, littered with special LaTeX-reserved characters like “#”, “_”, “%”, “&”, “@”, “~”, …etc.

The hyperref package apparently does some sophisticated gymnastics to handle the special chars. The \url{} and \href{}{} macros work for most (if not all) chars. But it becomes a shit-show when the same URL is used in multiple places in multiple representations. E.g. I often need to have a hyperlink as a readable string that visits an URL when clicked. \href from the hyperref pkg does that. But of course the URL is lost when the doc is printed. So the URL needs to become a footnote, which means the shitty-looking ungodly long URL must be entered twice. For example, a first attempt might look like this:

The \href{https://lemmy.sdf.org/c/tex_typesetting}{TeX community}\footnote{\scriptsize\verb|https://lemmy.sdf.org/c/tex_typesetting|} is where we discuss `\LaTeX`.

There is an underscore, which probably has to be escaped in the \verb but not the \href (not sure ATM- it’s hard to keep track of all the exceptions). Of course it’s quite annoying that the URL appears twice. So the temptation to thwart redundancy leads to this:

\newcommand{\hardlink}[2]{\href{#1}{#2}\footnote{\scriptsize#1}}

Disaster, because some chars need to be escaped for the footnote but if you escape those same chars for the \href, the backslashes appear literally in the hyperlink which is broken as a result. Another problem is when using minipages to get two columns, the footnote width is cut in half thus forcing long URLs to wrap at the horizontal midpoint of the page instead of continuing and using the wasted footer space under the right column. \mbox fixes that. So after much fiddling with blind hacks like \urlescape, \noexpandarg, \normalexpandarg, and \expandafter, I arrived at this:

\newcommand{\hardlink}[2]{\href{#1}{#2}\footnote{\scriptsize\mbox{\detokenize{#1}}}}

The \detokenize works for some special chars but not others. So still a fuckin’ mess. A LaTeX wizard of sorts went off to work on this problem for me, and came up with this:

\makeatletter
% NOTE: The following is an ugly hack that temporary redefines an internal
%       command of hyperref to process the verbatim URL. There is no warranty
%       and no support for this code or documents using this code!
\newcommand*{\footnotelink}{%
  \global\let\original@hyper@@link\hyper@@link
  \let\hyper@@link\onetime@special@hyper@@link% ugly hack
  \href
}
\newcommand*{\onetime@special@hyper@@link}[3]{%
  \global\let\hyper@@link\original@hyper@@link% ugly hack
  \hyper@@link{#1}{#2}{#3}%
  \IfArgIsEmpty{#2}{\footnote{\tiny\nolinkurl{#1}}}{\footnote{\tiny\nolinkurl{#1\##2}}}%
}
\makeatother

That monstrosity is the nuclear option that works in most cases. But IIRC it still fucks up in some situations, so I must use a combination of \hardlink and \footnotelink.

But what about QR codes? Fuck me. Another dimension of the same problem. Producing a doc with QR codes but not the URL strips the reader of some dignity. But a footnote is a bad way to expand a barcode. The URL should appear close to the QR code so the reader need not hunt for it. But URL size and circumstances ensure we cannot simply make a macro that hard codes it. Every layout situation is different.

Having a bibliography section helps force a standard presentation, but that still requires the URL to be repeated and we don’t necessarily want to be forced to have a bibliography anyway.

What we really need is an URL database, which maps tokens to URLs in the preamble. Consider how the datetime2 pkg works. You can store a list of dates like this:

\DTMsavedate{event1}{2021-12-10}
\DTMsavedate{event2}{2022-02-21}
\DTMsavedate{event3}{2022-03-10}
…
\begin{document}
yada yada \DTMusedate{event1} yada yada \DTMsetstyle{ddmmyyyy}\DTMusedate{event3}…
lorem ipsum \DTMsetstyle{mmddyyyy}\DTMusedate{event2} lorem ipsum \DTMsetdatestyle{default}\DTMsetup{datesep=/}\DTMusedate{event1} …etc.

We need that for URLs. Simply making a \newcommand for each URL would not work because \qrcode, \href, \texttt, \verb and family of verbatim envs all treat the special chars differently and some do not even expand commands. It needs to be a macro that can probe its own user to know which chars to escape.

One of the markdown languages supports URL references. E.g. you can declare ergonomic names for the URLs:

[diseasePlusCure]: https://krebsonsecurity.com/2016/10/spreading-the-ddos-disease-and-selling-the-cure
[diseasePlusCure-ia]: http://web.archive.org/web/20230713212522/krebsonsecurity.com/2016/10/spreading-the-ddos-disease-and-selling-the-cure
[mislabelling-ia]: <http://web.archive.org/web/20211006120915/people.torproject.org/~lunar/20160331-CloudFlare_Fact_Sheet.pdf#page=3>
[fediThreat]: https://write.pixie.town/thufie/dont-trust-cloudflare
[fediThreat-ia]: http://web.archive.org/web/20230827161847/write.pixie.town/thufie/dont-trust-cloudflare
[testamony]: https://dragonscave.space/@BlindMoon38/111954315299607397
[personalisedPricing]: https://web.archive.org/web/20240601161454/http://robindev.substack.com/p/cloudflare-took-down-our-website

Then in the doc write: “Cloudflare exploits [personalised pricing][personalisedPricing]” so the shitty URL does not obnoxiously pollute the text.

A LaTeX approach could be:

\savelink{CloudflareLies}{http://web.archive.org/web/20211006120915/https://people.torproject.org/~lunar/20160331-CloudFlare_Fact_Sheet.pdf#page=3}
…
\begin{document}
Yada yada.. This QR: \uselink[form=qr, width=20mm]{CloudflareLies} leads to \uselink[form=fixedwidthfont,wrapping=false]{CloudflareLies}.

We have something that partially works and is not well documented. There is an url package and there is a hyperref package. The hyperref package is said to supercede the url pkg. In fact, hyperref uses the url pkg. So you would generally ignore the url pkg and just use hyperref. But the docs for hyperref conceal the existence of the \urldef command from the url pkg. The docs for url show that you can do this:

\urldef{\nastyurl}\url{http://www.musical-starstreams.tld/~william@orbit/very_long_using_underscores/and%20spaces/file+with^caret.pdf?arg1=x&arg2=y#page=5}

Then you can use \nastyurl throughout your doc, including inside footnotes. But it screws up when used inside \href (no wonder hyperref docs neglect to mention it). It also falls over when used inside \qrcode.

My fight against enshitification entails hand-delivering paper correspondence by bicycle. I also re-use the windowed envelopes that I receive. When I run out those, I leave the envelope unsealed so the recipient can easily reuse the new envelope.

The recipient generally must use the postal service to reply. And rightly so. Penalizing them with the cost of printing and posting serves as a means to punish them for the enshitified digital path they offer. For me, this approach sufficiently casts my anti-enshitification votes while supporting the postal service that gives refuge from enshitification, without excessive environmental detriment.

The LaTeX scrlttr2 class is useful for using and re-using windowed envelopes. If the envelope is standard, the geometry may be known to the supplied KOMAscript machinery. If not, a few measurements can be given as parameters to align an address in a custom window.

To load the US №9 standard envelope, you would start with:

\documentclass[UScommercial9]{scrlttr2}

or for the French standard:

\documentclass[NF]{scrlttr2}

If you reuse a non-standard windowed envelope, you can put the following in the preamble and tamper with the measurements as needed:

\makeatletter
\setplength{foldmarkhpos}{4.2mm}   % default=3.5mm; distance from paper edge to fold mark; should account for the unprintable area of your printer
\setplength{tfoldmarkvpos}{108mm}  % default=99mm; distance between top fold mark and top paper edge
\setplength{firstheadwidth}{190mm} % default=170mm for NF and \paperwidth for others; width of letterhead
\setplength{firstheadvpos}{10mm}   % default=15mm for NF; distance from top edge to letterhead
\setplength{toaddrvpos}{40mm}      % default=35mm; distance between top of window and top paper edge
\setplength{toaddrhpos}{98mm}      % default=-10mm; distance from the left edge of the paper to the address field (if positive)
\setplength{toaddrindent}{5mm}     % default=10mm; left and right indentation of the address within the to-address box
\setplength{toaddrheight}{40mm}    % default=45mm
\makeatother

I’ve created a very customized LaTeX document which contains portions of machine translated text. I will ask a native speaker to make the text proper. I’m not sure who will take on the task yet, but it’s unlikely to be someone who understands LaTeX. My large preamble would add to the intimidation.

I think overleaf would have normally been ideal. But it became restricted access and hostile toward Tor users a few years ago. I cancel oppressive platforms like that.

One idea is to host it on some arbitary gitea server. They can probably edit the text directly in the web browser of they are a low-tech user. Or if tech-proficient they can use git as it was designed. Doesn’t matter if they butcher the code.. I’ll deal with the cleanup. I guess my main concern is that they would be so alienated by the code that it would put them off this volunteer effort.

Pandoc was one thought: pandoc -o paper.md -f latex -t markdown paper.tex, in which case they would work in a less alien situation. But pandoc can’t even handle my 2-column doc. It falls over on a tabular and produces nothing. But even if it could produce results, I’d expect disaster anyway.

Probably no great answers here.

High-level EU courts apparently assume all those who read their acronym-littered opinions and judgements are Subject Matter Experts (SMEs) who already know what the acronyms stand for.

I’m not a lawyer but this seems sloppy from a legal standpoint because an acronym that is never expanded is ambiguous. It creates room for confusion and misinterpretation in the worst case, and in the very least wastes the reader’s time on investigation.

Have lawyers and judges not been trained on this? As a technologist, my training included the good practice of expanding every single acronym the first time it appears, as I did above with “SME”, as well as the extra diligent but optional practice of including a section at the end with all expansions.

I realise that the whole legal industry is made up of mostly tech illiterates. Geeks have the advantage of being able to use LaTeX with the acro package¹, which enables us to write acronyms without thinking about where it first appears because the software automatically expands the first occurrances (or as we specify). Legal workers have probably limited themselves to dumbed down tools like MS Word which probably does not automate this, but nonetheless it’s the writer’s duty to see that acronym expansion happens.

Abbreviations:

SME: Subject Matter Expert

¹ In LaTeX, the preamble would have \DeclareAcronym{sme}{short=SME, long=Subject Matter Expert} and throughout the document each instance would be written as \ac{sme}.

It’s disturbing that infosec illiterate friends enter my name and contact details into their Android phonebooks, which then gets recklessly shared in countless ways outside of my control and without my knowledge or consent as to which data abusers ultimately get my contact info.

I try to practice data minimisation even with friends (if they are new), so I don’t give them an email address; generally just my first name, XMPP acct, and phone number. But then of course they enter my name into their dodgy phonebook along with my last name if they happen to get it circumstantially.

So I have a fix of sorts. We can have some control over how the info gets entered into people’s phonebooks by using a vCard. One option is to leave your name blank on the vCard but to graphically put your name in the avatar image on your vCard. OTOH, users will likely manually fill your name in anyway. So consider using the name field but deviating from normal text. You can find some obscure unicode fonts at yaytext.com. Then follow this LaTeX template to generate a contact card:

\documentclass{minimal}

\usepackage[paper=a4paper,layoutwidth=210mm,layoutheight=297mm]{geometry}
\usepackage[newdimens]{labels}% let the package do the work...
\usepackage{qrcode}
 
% These attributes are for European label sheet OLW4738
\LabelCols=3
\LabelRows=7
\LeftPageMargin=0in
\RightPageMargin=0in
\TopPageMargin=0in
\BottomPageMargin=0in
\InterLabelColumn=0mm% adjust as required
\InterLabelRow=0mm
\RightLabelBorder=0mm% adjust to taste
\LeftLabelBorder=0mm
\TopLabelBorder=2mm
\BottomLabelBorder=2mm

\LabelGridtrue % <== use to line stuff up; delete this line to process final version
\numberoflabels=12 % ← normally this is 21 to fill a page (3×7), but due to memory overflow bug w/too many QR codes, it must be reduced!

\begin{document}

\genericlabel{%
  \begin{minipage}{66mm}% actual label is 70mm wide; subtract \RightLabelBorder and \LeftLabelBorder
    \hspace*{4mm}%
    \qrcode[height=22mm, level=l]{BEGIN:VCARD\?
VERSION:4.0\?
N:刀囗モ;╝ǫⱨᶇ;;;\?
IMPP:xmpp:johnsnickname@jd.snikket.chat\?
TEL;VALUE=uri;TYPE="cell":tel:+①-𝟝𝟝𝟝-𝟝𝟝𝟝-①²①²\?
LANG:en\?
END:VCARD
}%
\parbox[c]{8em}{%
  snkt fingrprint $\rightarrow$\\
  \vfill
  $\leftarrow$ Vcard4\\
  \vfill
  dino fingerprint $\rightarrow$
}
\parbox[c]{11mm}{
  \qrcode[height=11mm, level=l]{xmpp:johndoe@jd.snikket.chat?omemo-sid-1234567890=a9a9dc175fbdebad99db71f72396a1e7a9a9dc175fbdebad99db71f72396a1}\\
  \vfill
  \qrcode[height=11mm, level=l]{xmpp:johndoe@jd.snikket.chat?omemo-sid-1234567890=75fbdebad99db71f72396a1e7a9a9dc175fb1e7a9a9dcfbdebad99db71f723}
}
\end{minipage}
}

\end{document}

It’s not infallible but it’s unlikely that enough people would be doing this to justify Google coding their identity cross referencing logic to decode atypical characters.

It’s not trivial to get a good font. A lot of the yaytext fonts are simple font changes so when the QR code is scanned, the phone seems to automatically convert the font back to normal characters. Unfortunately this means you need to carefully select a non-font style that is being abused as a font, which then leads your name to look like a ransom letter style.

Kids can use cool nicknames w/out a real name to mitigate the problem to some extent, especially if they’re a hipster drug dealer, but it’s harder for an adult to pull that off without alienating people and coming off as a kid. We need to at least try to pretend to behave like adults.

It would be nice if there were a desktop app that could give all the yaytext.com styles and a bit more of the obscure ones. There is some python code in this thread but it’s quite limited in fonts. It’s missing the good ransom letter fonts.

(I tried to cross-post to cybersecurity@infosec.pub but this post triggers the slur filter there so I could not post it.)

LaTeX is great for writing letters. It seems like a little known secret how well the scrlttr2 class formats letters for windowed envelopes. LaTeX really makes letter writing enjoyable for programmers (though it would likely be hell for non-programmers).

If I were using a WYSIWYG tool like Libre Office, writing letters would be mundane, boring, and tedious. And the results would be aesthetically limited without doing copious manual labor.

There is noteworthy gratification in turning letter writing into a programming exercise. So whenever a gov agency or corporation fucks me over in some way, I find it theraputic to write complaints and petitions in LaTeX.

There is an hacktivist mantra that goes something something like this:

“write code not text” (not sure on the exact wording)

LaTeX basically turns that on its side because you do both at the same time. I have built up a library of captioned legal statutes in LaTeX, such as commonly referenced GDPR law. So I can crank out GDPR requests quite quickly by using \input statements that imports a very nicely formatted block quote of law which I have thoroughly over-engineered. Also fun to use the qrcode package to reference URLs.

The perfectionism probably consumes more time than using a shit tool like MS Word in the end. But it’s enjoyable. And because it’s enjoyable, it triggers writing more petitions and complaints that I would otherwise write. Every time I get fucked over by some administrative malpractice, it’s another fun opportunity to play in LaTeX and refine my code.

I was disappointed to see that the qrcode package gives no way to insert an image into the center of the QR code. But in fact it turns out that QR codes cannot be made to have an alternate center. QR codes are simply spec’d to have 30% redundancy. So you can simply overwrite up to 30% of a QR code arbitrarily and it will still decode as long as you don’t mess with the boxes on the 4 corners.

Also worth noting that you can exceed 30% interference if you play games with colors. That is, if a transparent pic uses sufficiently light colors that pass as white (in a black vs white dithering algo), then those pixels obviously don’t count in the 30% tolerance. So some quite clever work could exploit this to make a QR code look less like a pixel blob.

I guess the gripe that I have is that redundancy is fixed at 30% for all QR codes, IIUC.

In principle, we should be able to generate a code with 50% redundancy and then clobber up to 50% of it.

It’s a slightly labor intensive because for each line of text you have to specify an endpoint.. but it’s managable enough. Also worth considering is Inkscape, which has a function to flow text into a shape.

It would be fun to collect some templates for re-use. E.g. if someone wants to complain about the corrupt tyrant who just took power (the most powerful office in the world) a couple days ago, a middle finger would be appropriate for that sort of thing.

Suppose you feed a multi-page PDF into \includepdf. If you only want pagecommand or picturecommand to take effect on some pages, you normally must split the construct up into multiple invocations. E.g.

\includepdf[pages=1], pagecommand={\doStuffOnPageOne}]{file.pdf}
\includepdf[pages=2-], pagecommand={\doStuffOnPagesAfterOne}]{file.pdf}

It gets ugly fast when there are some commands you want performed on every page, and some on select pages, because then you must write and maintain redundant code.

Fuck that. So here’s a demonstration of how to write code inside pagecommand and picturecommand that is page-specific:

\documentclass{article}

% Demonstrates use of the pdfpages package with page-by-page actions that are specific to select pages.

\usepackage{mwe}        % furnishes example-image-a4-numbered.pdf
\usepackage{pdfpages}
\usepackage{pdfcomment}
\usepackage{ocgx2}      % furnishes the ocg environment (PDF layers, but not really needed for this demo)
\usepackage{fancybox}   % furnishes oval box
\usepackage{fontawesome} % furnishes \faWarning

\begin{document}

\makeatletter
\includepdf[pages=1-,pagecommand={%
  % \makeatletter ← ⚠ does not work in this scope; must wrap the whole includepdf construct
  \pdfcomment[icon=Note, hoffset=0.5\textwidth, voffset=5em]{%
    (inside pagecommand, executing on every page)\textLF\textLF
    \texttt{\textbackslash AM@page} variable: \AM@page\textLF\textLF
    Side-note: the voffset option has no effect when the value is positive (5em in this case)% 
  }%
  \ifthenelse{\AM@page = 1 \OR \AM@page = 2 \OR \AM@page = 12}{%
    \pdfcomment[icon=Insert, hoffset=0.5\textwidth, voffset=-6em]{%
      (inside pagecommand, affecting only pages 1, 2, and 12)\textLF\textLF
      \texttt{\textbackslash AM@page} variable: \AM@page\textLF\textLF
      Strangely, the voffset option only works if it is negative.% 
    }%
  }{}
  % \makeatother
}, picturecommand={%
  \put(50,715){Inside the picture environment:}
  \put(50,700){%
    \begin{tabular}[t]{llp{0.6\textwidth}}
      internal \texttt{\textbackslash @tempcnta}    variable (useless):     &\the\@tempcnta&\\
      internal \texttt{\textbackslash @tempcntb}    variable (useless):     &\the\@tempcntb&\\
      internal \texttt{\textbackslash AM@pagecnt}   variable (useless):     &\the\AM@pagecnt&\\
      internal \texttt{\textbackslash AM@abs@page}  variable (useless):     &\AM@abs@page&\\
      internal \texttt{\textbackslash AM@page}      variable (interesting): &\AM@page & \faWarning Inside picturecommand, this number is 1 higher than the actual page number! But it’s correct inside pagecommand (see the annotation note to check).\\
      internal \texttt{\textbackslash AM@pagecount} variable (interesting): &\AM@pagecount&\\%
    \end{tabular}
    % lastpage: \AM@lastpage% broken
    \ifAM@firstpage
    We might expect this to trigger on the 1st page, but it never does. Likely because the page counter is incremented before picturecommand is invoked. It would perhaps work in the pagecommand construct.
    \fi
  }
  \put(500,770){% The ocg environment is irrelevant and unnecessary.. just here to demo PDF layers.
    \begin{ocg}{section labels}{sl1}{on}\color{blue}
      \Large\rotatebox{-45}{\setlength{\fboxsep}{6pt}\Ovalbox{Section~A}}
    \end{ocg}}}]%
{example-image-a4-numbered.pdf}
\makeatother
\end{document}

It would sometimes be useful to write conditional code that depends on boolean values defined in a parent package. E.g. the \pdfcomment package has the boolean “final”, which disables all PDF annotations in the document (\usepackage[final]{pdfcomment}). There is some other logic in my document that should also be disabled when that boolean is true. I tried simply using:

\ifpc@gopt@final\else%
…code that should not run when final is true…
\fi

pdflatex gives: “Undefined control sequence”

More generally, many draft options are often useful for controlling logic within the document for which a parent uses a draft option. Also when defining a custom letterhead in the scrlttr2 class there are booleans for many items that may or may not be wanted in the letterhead.

Has anyone managed to read a parent boolean?

(update) This thread gives useful options for many situations. But it does not completely answer the question because there are non-draft related booleans.

SOLVED

The \ifpc@gopt@final is reachable but only inside a \makeatletter stanza. Thus:

\makeatletter
\ifpc@gopt@final\else%
…code that should not run when final is true…
\fi
\makeatother

Some might find it useful to import a text file and put the contents into a PDF annotation. E.g. a PDF is in language A and you want to make a translation available in language B, in a PDF annotation.

Here’s how:

\documentclass{article}

\usepackage{mwe}
\usepackage{pdfpages}
\usepackage{pdfcomment}
\usepackage{newfile}
\usepackage{xstring}
\usepackage{catchfile}

% heredoc holding text that normally breaks the \pdfcomment command:
\begin{filecontents*}{\jobname_sample.txt}
line one

line two
tricky symbols: _&%
\end{filecontents*}

% normally the above file is whatever you supply to be imported into the PDF annotation. The heredoc is just to provide a self-contained sample.

% Create \pdfcommentfile, which is a version of \pdfcomment that can read from a file:
\makeatletter
\gdef\pdfcommentfile#1{%
  \begingroup
  \everyeof{\noexpand}%
  \long\edef\temp{\noexpand\pdfcomment{\@@input{#1}}}%
  \temp
  \endgroup
}%
\makeatother

\CatchFileDef{\cfile}{\jobname_sample.txt}{} % side-effects: replaces blank lines with “\par” and drops percent symbols

% Replace blank lines with \textLF and replace special symbols with those that are safe for \pdfcomment. Warning: this is probably not a complete list of all constructs that break \pdfcomment!
\StrSubstitute{\cfile}{\par}{\string\noexpand\string\textLF\ }[\pdfannotationtxt] % the hard space is after textLF is a bit unfortunate; not sure how to do a normal space there
\StrSubstitute{\pdfannotationtxt}{\%}{\string\noexpand\string\%}[\pdfannotationtxt]
\StrSubstitute{\pdfannotationtxt}{_}{\string\noexpand\string\_}[\pdfannotationtxt]
\StrSubstitute{\pdfannotationtxt}{&}{\string\noexpand\string\&}[\pdfannotationtxt]

% the \pdfcomment command cannot directly handle the above substitutions (nor can it handle the original unsubstituted version). So we write the new version to another file:

\newoutputstream{filteredresult}
\openoutputfile{\jobname_filtered.txt}{filteredresult}
\addtostream{filteredresult}{\pdfannotationtxt}
\closeoutputstream{filteredresult}

\begin{document}
\pdfcommentfile{\jobname_filtered.txt}
\includepdf{example-image-a.pdf}
\end{document}

There should be a way to substitute the special characters and blank lines then feed it directly to \pdfcomment, but I’ve exhausted that effort. I’ve been in this LaTeX rabbit hole for days now trying to do something that should be simple. So this is as far as I go. The code above works but it’s ugly as fuck that we have to write the filtered text to file then read the file back in. The file i/o slows down compilation much more than what I consider reasonable.

cross-posted from: https://linkage.ds8.zone/post/363360

I am trying to do some simple character replacements on an input file and writing it to a file. The output produced by \StrSubstitute is quite bizarre. Here is a MWE:

\documentclass{article}

\usepackage{newfile}      % furnishes \newoutputstream
\usepackage{catchfile}    % furnishes \CatchFileDef
\usepackage{xstring}      % furnishes \StrSubstitute
\usepackage{stringstrings}% furnishes \convertword (a \StrSubstitute alternative)

% heredoc that creates source input file
\begin{filecontents*}{\jobname_sample.txt}
line one

line two
tricky symbols: _&%
\end{filecontents*}

\CatchFileDef{\cfile}{\jobname_sample.txt}{}

\begin{document}

% Replacements needed:
%   & → \&
%   % → \%
%   _ → \_
%   \newline\newline → \textLF (replace blank lines)
%
\StrSubstitute{\cfile}{&}{\&}[\mystring]
\StrSubstitute{\mystring}{\%}{\%}[\mystring]
\StrSubstitute{\mystring}{_}{\_}[\mystring]
\StrSubstitute{\mystring}{\newline\newline}{\\textLF}[\mystring]

\newwrite\myoutput
\immediate\openout\myoutput=\jobname_filtered_native.txt
\immediate\write\myoutput{\mystring}
\immediate\closeout\myoutput

\newoutputstream{filtered}
\openoutputfile{\jobname_filtered_newfile.txt}{filtered}
\addtostream{filtered}{\mystring}
\closeoutputstream{filtered}

\noindent\textbf{filtered catchfile}:\\
\mystring

\noindent\textbf{filtered catchfile (2nd attempt)}:\\
\convertword{\mystring}{\newline\newline}{\noexpand\textLF}
 
\end{document}

That uses two different techniques to write to a file, and both give slightly different yet wildly unexpected output:

$ cat sample_code_filtered_native.txt
line one \par line two tricky symbols: \protect \global \let \OT1\textunderscore \unhbox \voidb@x \kern .06em\vbox {\hrule width.3em}\OT1\textunderscore \&
$ cat sample_code_filtered_newfile.txt
line one \par line two tricky symbols: \global\let \OT1\textunderscore \unhbox \voidb@x \kern .06em\vbox {\hrule width.3em}\OT1\textunderscore \&

What triggered all that garbage to be created? This is what the output [b]should[/b] be:

line one\textLF
line two
tricky symbols: _&%

I also tried a 3rd way to write \mystring to a file, as follows:

\begin{filecontents*}{\jobname_myvar.txt}
  \mystring
\end{filecontents*}

That approach literally writes the string “\mystring” to a file, which is useless in this case.

(update) apparently a \string needs to prefix the substituted strings.

Took me longer than I'd like to admit to realize that \directlua is first expanded before it goes into the lua interpreter, and that \% is defined through \chardef (in plain), which means that it's not expandable.
Luckily LuaTeX has the \csstring primitive.
Is anyone else doing any fun things with \directlua?

I’m just getting into LaTeX and am starting with a project I’ve cloned from GitHub. I immediately ran into problems compiling because of a bunch of missing packages. I was able to get it running by compiling, seeing where it failed, and installing the missing package, but I had to do this one at a time for over a dozen packages. Is there any sort of requirements.txt or package.json file that lists all dependencies so I can pipe them to the package manager to install?

What's your method for dealing with underfull/overfull \hboxes and unacceptable badness in general?

LaTeX has the \sloppy command which IIRC sets \tolerance to 9999, and \emergencystretch to a large value. But the default \tolerance is 200 (I think), which is a big difference. It's very "either/or" when maybe there's a more optimal way.

For my native language (swedish) I've found that many issues arise because TeX doesn't find all the possible hyphenation points, so I usually spend time adding words to the hyphenation list.

But still, in any longer text there's usually a couple of paragraphs that just won't set right, I'm curious about your tricks or methods for dealing with them.

ConTeXt has a nice font selection system. LaTeX ported the ConTeXt code in luaotfload.sty and they have fontspec, OpTeX has a font selection system with font files and all. In plain, there's only the primitive font switches. There are some packages on ctan for plain to extend functionality, but I'm not sure how they work.

The good news is, you can use luaotfload.sty directly in plain! Just \input luaotfload.sty. The bad news if you're into minimalism is that it depends on LaTeX so you'll need that installed. An alternative is to use luafonts.tex from the csplain package: \input luafonts.tex, it uses the luaotfload code too.

Once you've done that, you can use all the nice things in luaotfload.

In this example I'll use an updated version of Tuftes Bembo-clone, ETbb. You can put the files anywhere where luaotfload will find them, ~/.fonts or your projects directory for example.

There are many ways to implement font selection. I rarely use many fonts in a project, so I usually just do something simple like this:

\font\tenrm "ETbb-Regular" at 10pt
\font\tenit "ETbb-Italic" at 10pt
\font\tenbf "ETbb-Bold" at 10pt
\font\tenbi "ETbb-BoldItalic" at 10pt
\font\tensc "ETbb-Regular":+smcp;letterspace=10; at 10pt
\font\tencaps "ETbb-Regular":+upper;letterspace=10; at 10pt

The opentype features come after the name, with a + or - to turn them on or off. To make it a little more semantic I add a size macro, and why not set \baselineskip at the same time. You could also set struts here.

\def\normalsize{% 10pt
	\baselineskip=12pt
	\def\rm{\tenrm}%
	\def\it{\tenit}%
	\def\bf{\tenbf}%
	\def\bi{\tenbi}%
	\def\sc{\tensc}%
	\def\caps{\tencaps}%
}

Now I can type \normalsize\rm and the default will be 10pt roman. \it will switch to italic, \sc to small caps, etc. I have two special switches for small caps and big caps because I always want them letterspaced and maybe some opentype features too.

With the same structure, it's (repetitive) but easy to add a \footnotesize say, 8pt, and a \largesize at 12pt.

In regular writing the macros could work something like this:

\normalsize\rm % default for document

\centerline{\largesize\caps Title}

\vskip\baselineskip

Lorem ipsum {\it dolor} sit amet, consectetur {\bf adipiscing} elit. Integer non {\bi accumsan} sem. Vestibulum ante {\sc ipsum} primis in faucibus orci luctus et ultrices posuere cubilia curae; Morbi blandit in nisl sed dapibus. Praesent porttitor id mauris sit amet tincidunt.
 
\vskip\baselineskip 
 
{\footnotesize\rm
Lorem ipsum {\it dolor} sit amet, consectetur {\bf adipiscing} elit. Integer non {\bi accumsan} sem. Vestibulum ante {\sc ipsum} primis in faucibus orci luctus et ultrices posuere cubilia curae; Morbi blandit in nisl sed dapibus. Praesent porttitor id mauris sit amet tincidunt.\par
}

which gives:

This is a very primitive and simple way, and would probably become tedious if you're using lots of different fonts, then it would be better to use/make a more advanced system. There's a programming paradigm called "Worse is better", and I'm not sure if this is an example of that. Maybe it's just "Worse is worse". But, it's easy to understand all the moving parts, which can be a good thing.

The full code:

%\input luaotfload.sty
\input luafonts.tex

\hsize=65mm
\frenchspacing
\tolerance=1000

\font\eightrm "ETbb-Regular" at 8pt
\font\eightit "ETbb-Italic" at 8pt
\font\eightbf "ETbb-Bold" at 8pt
\font\eightbi "ETbb-BoldItalic" at 8pt
\font\eightsc "ETbb-Regular":+smcp;letterspace=10; at 8pt
\font\eightcaps "ETbb-Regular":+upper;letterspace=10; at 8pt

\font\tenrm "ETbb-Regular" at 10pt
\font\tenit "ETbb-Italic" at 10pt
\font\tenbf "ETbb-Bold" at 10pt
\font\tenbi "ETbb-BoldItalic" at 10pt
\font\tensc "ETbb-Regular":+smcp;letterspace=10; at 10pt
\font\tencaps "ETbb-Regular":+upper;letterspace=10; at 10pt

\font\twelverm "ETbb-Regular" at 12pt
\font\twelveit "ETbb-Italic" at 12pt
\font\twelvebf "ETbb-Bold" at 12pt
\font\twelvebi "ETbb-BoldItalic" at 12pt
\font\twelvesc "ETbb-Regular":+smcp;letterspace=10; at 12pt
\font\twelvecaps "ETbb-Regular":+upper;letterspace=10; at 12pt

\def\footnotesize{% 8pt
	\baselineskip=10pt
	\def\rm{\eightrm}%
	\def\it{\eightit}%
	\def\bf{\eightbf}%
	\def\bi{\eightbi}%
	\def\sc{\eightsc}%
	\def\caps{\eightcaps}%	
}

\def\normalsize{% 10pt
	\baselineskip=12pt
	\def\rm{\tenrm}%
	\def\it{\tenit}%
	\def\bf{\tenbf}%
	\def\bi{\tenbi}%
	\def\sc{\tensc}%
	\def\caps{\tencaps}%
}

\def\largesize{% 12pt
	\baselineskip=14pt
	\def\rm{\twelverm}%
	\def\it{\twelveit}%
	\def\bf{\twelvebf}%
	\def\bi{\twelvebi}%
	\def\sc{\twelvesc}%
	\def\caps{\twelvecaps}%
}

{\footnotesize\rm footnotesize rm}\par
{\footnotesize\it footnotesize it}\par
{\footnotesize\bf footnotesize bf}\par
{\footnotesize\bi footnotesize bi}\par
{\footnotesize\sc footnotesize sc}\par
{\footnotesize\caps footnotesize caps}\par

\vskip\baselineskip

{\normalsize\rm normalsize rm}\par
{\normalsize\it normalsize it}\par
{\normalsize\bf normalsize bf}\par
{\normalsize\bi normalsize bi}\par
{\normalsize\sc normalsize sc}\par
{\normalsize\caps normalsize caps}\par

\vskip\baselineskip

{\largesize\rm largesize rm}\par
{\largesize\it largesize it}\par
{\largesize\bf largesize bf}\par
{\largesize\bi largesize bi}\par
{\largesize\sc largesize sc}\par
{\largesize\caps largesize caps}\par

\vskip\baselineskip 
\hrule
\vskip\baselineskip

\normalsize\rm

\centerline{\largesize\caps Title}

\vskip\baselineskip

Lorem ipsum {\it dolor} sit amet, consectetur {\bf adipiscing} elit. Integer non {\bi accumsan} sem. Vestibulum ante {\sc ipsum} primis in faucibus orci luctus et ultrices posuere cubilia curae; Morbi blandit in nisl sed dapibus. Praesent porttitor id mauris sit amet tincidunt.
 
\vskip\baselineskip 
 
{\footnotesize\rm
Lorem ipsum {\it dolor} sit amet, consectetur {\bf adipiscing} elit. Integer non {\bi accumsan} sem. Vestibulum ante {\sc ipsum} primis in faucibus orci luctus et ultrices posuere cubilia curae; Morbi blandit in nisl sed dapibus. Praesent porttitor id mauris sit amet tincidunt.\par
}

\bye

This is a rather interesting presentation by Jean-Luc Doumont about placing content on a grid, or a subset of the grid. He's kind of taking this idea to the extreme.

The defaults for headline and footline in plain.tex is:

\headline={\hfil}
\footline={\hss\tenrm\folio\hss}

I.e. there's no header, and in the footer the page number is printed in the center.

Many times I've found that some pages, like title pages or chapter opening pages, need exceptions from the defaults. I'd like to have a way of saying "on this page there should be neither header nor footer" or "on this page there should be a footer, but no header". Or, the header should say one thing if it's a verso page, and another if it's a recto page. In this case, we can use conditionals. They can be created like this:

\newif\iffootline \footlinetrue
\newif\ifheadline \headlinetrue

I want the default to be to print both headline and footline, so the conditionals are set to true initially.

In this case I want to control the current page only, so if the conditional is false, it needs to change back to true for the next page. For the footline, we could have this:

\footline{%
	\iffootline
		\hss\tenrm\folio\hss
	\else
		\hfil
		\global\footlinetrue	
	\fi
}

If the conditional is true, the footline is to be printed. If it's false, it prints nothing, but switches back the conditional to true for the next page.

We can do the same thing with the headline, but also add different headline texts if it's an odd or even page:

\headline{%
	\ifheadline
		\ifodd\pageno
			\hss\tenit Title\hss
		\else
			\hss\tenit Author\hss
		\fi
	\else
		\hfil
		\global\headlinetrue
	\fi
}

These conditionals can then be used in other macros, like for chapter headings, or used as they are by placing \headlinefalse and/or \footlinefalse on pages where you don't want headers and/or footers.

It's been around... looks at clock ...nine years since something significant happened with the EB Garamond initials, but a couple of months ago they split off into their own github repo and you can now use all the letters of the alphabet! Before, it was limited to a subset of letters.

NB, this only applies to the original Duffner EB Garamond, not the Octavio Pardo version found on Google Fonts.

If you want them, clone or download the repo, download the dependencies, and run make. Or message me and I'll send the files. It's free software so that ought to be ok.

There's this guy called Stephan V. Bechtolsheim who wrote a series of books on TeX called "TeX in Practice". He's really good at the finer details of, well, everything.

In one of the books, he makes a macro to visualize boxes. It's built up over many pages and chapters with lots of small macros as building blocks. Because he's reusing these macros in different places, it makes a lot of sense.

However, when I wanted to use the box visualizing macro, I found that I had to look up and copy a lot of code to make it work. This was no fun, so I re-wrote it in a "flatter" way where it's just regular plain old macros.

I ended up with this:

\newdimen\linethickness \linethickness=0.4pt

\def\boxlines #1{%
    \hbox{%
        % Save original (argument) box
        \setbox0 = #1%
        % Place bullet at baseline and vertical align of the box
        \setbox1 = \hbox{\hskip -2.5pt \lower 2.5pt \hbox{$\circ$}}%
        \ht1=0pt \dp1=0pt \wd1=0pt
        \box1
        % Place a dashed line at baseline
        \setbox2 = \hbox to \wd0{%
            \xleaders\hbox to 4pt{%
                \hskip 1pt
                \vrule depth 0.5\linethickness 
                	   height 0.5\linethickness 
                	   width 2pt
                \hfil
            }%
            \hfil
        }%
        \ht2=0pt \dp2=0pt \wd2=0pt 
        \box2
        % Place frame
        \setbox 3 = \hbox{%         
            \hskip -0.5\linethickness
            \vrule width \linethickness height \ht0 depth \dp0% 
            \hskip \wd0% 
            \hskip -\linethickness
            \vrule width \linethickness height \ht0 depth \dp0% 
            \hskip -\wd0% 
            \hskip -\linethickness
            \dimen0 = \wd0% 
            \advance\dimen0 by \linethickness
            \dimen2 = \ht0% 
            \advance\dimen2 by 0.5\linethickness
            \dimen4 = \ht0% 
            \advance\dimen4 by -0.5\linethickness
            \dimen4 = -\dimen4
            \vrule width \dimen0 height \dimen2 depth \dimen4
            \hskip -\dimen0
            \dimen2 = \dp0% 
            \advance\dimen2 by -0.5\linethickness
            \dimen2 = -\dimen2
            \dimen4 = \dp0% 
            \advance\dimen4 by 0.5\linethickness
            \vrule width \dimen0 height \dimen2 depth \dimen4
        }%
        \ht3=0pt \dp3=0pt \wd3=0pt 
        \box3
        % Place original argument box
        \box0
    }%
}

The macro takes a box as an argument, for example \boxlines{\box0}

It puts lines around the box, and marks out the baseline and the horizontal alignment. The neat thing is that the lines are made so that they don't interfere with the typesetting at all. Everything is placed as it would be without the lines.

If you have something that looks misaligned or strange, like these words:

it can help to visualize the boxes:

Sometimes it's nice to have a macro that prevents indentation of the next paragraph. For example, if you're writing your own section header macros and it's hard to automatically insert a \noindent in the right place. There's another way but it's not very intuitive:

\def\firstnoindent{%
    \global\everypar={%
        \global\everypar={}%
        \setbox0=\lastbox
    }%
}

When I first came across this I had to pause and think about it a little.

What happens when you start a new paragraph? First you're in vertical mode, then TeX interprets some tokens as horizontal material and wants to switch over to horizontal mode. \noindent is a special horizontal material that starts an empty horizontal list, but all other material, like regular characters or horizontal glue, will first insert an \hbox of \parindent length. This is your normal paragraph indentation. Next, TeX will insert the content of \everypar, and then come your words or whatever made TeX start a paragraph. This order is important. \hbox - \everypar - .

When you place \firstnoindent somewhere, it sets \everypar to contain two things. 1) An instruction to set \everypar to be empty. This makes sure that only the next paragraph is affected. 2) It sets \box0 to contain the last box created, which makes the last box disappear.

So this is now in \everypar, but it won't be expanded until a paragraph is created.

When the next paragraph is created, TeX inserted the indentation \hbox, then the content of \everypar, which wipes \everypar and does the other thing: \setbox0=\lastbox.

\lastbox is a primitive that removes the last box, then immediately inserts it again. By itself it changes nothing. The idea is that you can make it disappear as the input to a \setbox, do things to that box, then manually put it back with \box. But here we don't put it back, it just disappears. And the last created box was the indentation \hbox.

So, after putting \firstnoindent somewhere, we'll make the next indentation disappear, and the next paragraph after that \everypar will be empty, and will indent as usual.

Caveat 1: This assumes that \everypar is normally empty.

Caveat 2: I'm on my phone and haven't tested the code, but it should work, in theory, maybe.

I've had projects where it's just one big .tex file including all the macros and content, and projects where there's a whole file hierarchy of chapter subdirectories and layers of \inputs and macro files and... It can sort of get out of hand for me, where it makes sense in the moment and then looking at it a week later it's a puzzling labyrinth of files.

I guess it depends on the nature of the project, but do you have a go-to way that works for you in most cases?

There's an example of a dropcap in the TeXbook but it invokes \lineskip because of the big letter, and so makes it hard to stay on the grid. Another solution is to first skip down a few lines, place a big letter in a box of zero height, and then skip back up again and shape the paragraph with \parshape.

It could look something like this:

\font\tenrm "EBGaramond12-Regular" at 10pt \tenrm
\font\tensc "EBGaramond12-Regular":+smcp; at 10pt
\font\quotefont "EBGaramond12-Regular" at 23pt
\font\dropcapfont "EBGaramond12-Regular" at 46pt

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut ut cursus ipsum, feugiat accumsan mi. Pellentesque sed ante ac augue tincidunt ultricies. Mauris vehicula ipsum est, nec bibendum odio semper sit amet. Aenean fringilla purus est, ut varius enim consectetur non.

\vskip2\baselineskip
\smash{%
    \hskip0pt % For moving the cap into the margin.
    \raise 15pt \llap{\quotefont “\hskip0pt}%
    {\dropcapfont N}%
}%
\vskip-3\baselineskip
\dimen0=35pt % First line indent
\dimen1=38pt % Second line indent
\dimen2=38pt % Third line indent
\parshape 4
    \dimen0 \dimexpr(\hsize-\dimen0)
    \dimen1 \dimexpr(\hsize-\dimen1)
    \dimen2 \dimexpr(\hsize-\dimen2)
    0pt \hsize
\noindent
{\tensc ullam,” auctor quam eget porta feugiat.} Suspendisse mattis fringilla turpis, vel hendrerit orci hendrerit nec. Donec vel nisi vestibulum, placerat leo in, tempus nisi. Sed non lacus id velit accumsan ullamcorper. Nulla eget risus ex. Pellentesque faucibus sit amet orci a fermentum. Quisque finibus turpis quis elit lobortis, sed viverra erat eleifend. Nulla imperdiet felis ac augue fermentum, in tincidunt nibh pellentesque.

If you have fancy dropcaps with foreground and background like in EB Garamond you can layer the caps with an \rlap to make something like this:

As Knuths TeX had no support for colors or pictures, we have to make our own macros for that. The following macros are perhaps a bit simple, but that also means being easy to adapt to different needs and situations.

Images can be placed with an \image{} macro. It can be preceeded with \imageheight or \imagewidth to set dimensions. If a dimension is set to 0pt, the natural size of the image is used. This is very similar to how it's done in Opmac/OpTeX.

\newdimen\imagewidth \imagewidth=0pt % 0pt gives natural size of image
\newdimen\imageheight \imageheight=0pt
\newtoks\imagedir \imagedir{} % Must end with /, e.g. mypics/

\def\image#1{%
	\hbox{%
		\saveimageresource 
			\ifdim\imagewidth=0pt \else width\imagewidth\fi 
            \ifdim\imageheight=0pt \else height\imageheight\fi 
            {\the\imagedir#1}%
      	\useimageresource\lastsavedimageresourceindex
	}%
}

% Example: {\imageheight=4cm \image{mypicture.png}}

Colors can be set like this:

% Token for \aftergroup to reset color
\def\colorstackpop{\pdfextension colorstack 0 pop}

% Switch to CMYK color in current group. Expects four values between 0 and 1
\def\setcmykcolor#1{%
   	\pdfextension colorstack 0 push {#1 k #1 K}% 
   	\aftergroup\colorstackpop
}

% Switch to RGB color in current group. Expects three values between 0 and 1
\def\setrgbcolor#1{%
   	\pdfextension colorstack 0 push {#1 rg #1 RG}% 
   	\aftergroup\colorstackpop
}

% Color names
\def\black{\setcmykcolor{0 0 0 1}}
\def\blue{\setcmykcolor{1 1 0 0}}
\def\red{\setrgbcolor{0.8 0 0}}

% Example: {\blue This is blue text}