This repository contains the source files for every document, example, exercise and their solutions used as materials for the corporate trainings I provided on programming topics.
Last successful builds for each training:
 |
Combined documents (PDF) | All materials (zip) | Presentation slides (zip) |
|---|---|---|---|
| CDU (CUDA) | CDU_documents_combined.pdf | CDU_materials.zip | CDU_presentation_slides.zip |
| RUX (Rust) | RUX_documents_combined.pdf | RUX_materials.zip | RUX_presentation_slides.zip |
| THO (Python) | THO_documents_combined.pdf | THO_materials.zip | THO_presentation_slides.zip |
The most reliable way to build everything is to use the nix
package manager (with flakes enabled). Use either nix build to
build everything, or nix build .#CDU / nix build .#RUX / nix build .#THO to build one specific
training's materials.
Then go grab a coffee, that thing is so overengineered that it takes a while to build everything (see the sections below for how overengineered we're talking about).
After nix is done, you'll find the materials in the result symlink created in your current
directory.
For each training, nix will produce:
result/XXX_materials/CM*-handout.pdf: some PDF slide deck in a "handout" format with reading notes mimicking what I would say during its presentation;result/XXX_materials/examples/: the buildable and testable code files used as examples in the slides;result/XXX_materials/TP-*.pdf: a PDF practical with exercises;result/XXX_materials/given_files/,result/XXX_materials/exercises/, or similar: the code files given as a starting point for the exercises;result/XXX_materials/ref_files/,result/XXX_materials/correction/,result/XXX_materials/exercises_ref/, or similar: the code files with the solutions to the exercises;- occasionnaly other documents (such as the CDU training's "memo");
result/XXX_materials.zip: a zip file with all the above materials;result/XXX_presentation_slides/CM*.pdf: a folder containing only the presentation slides (with more transitions and without reading notes).
To build the PDF documents without nix, you will need the training's subject language toolchain
(except for the CDU training which does not actually need CUDA to build the documents), and a LaTeX
distribution with: latexmk,
xelatex, and the packages listed in the latex-distribution
definition of the flake.nix file. You can then go to a training's documents folder
and run latexmk to build all slides, practicals and memos, or latexmk <document_name> to build a
specific document.
To build the exercises and their solutions (which for some trainings need some processing to be
usable in a training session), the process depends on the training (sorry). Go to the training's
exercises folder (CDU/filters, RUX/TP, etc.) and use the provided Makefile or script (like
RUX's build_given_files.zsh). Please be advised that some of these builds will try to use nix to
load a development environment defined in the toplevel flake.nix file (which could involve
downloading most of the things that a global nix build would download).
This repository features no less than
- nix, to define the development and build environment;
- latexmk, to build all PDF documents (and run Python/Rust code examples, more on that later);
- make, to organize the built documents (some organization steps at least, nix then does the rest);
- some standalone shell scripts;
Some shell scripting is embedded in the nix flake definition, but the funniest part is in the
LaTeX documents themselves: common/latex/myhighlighting.sty
makes heavy use of VerbatimeOutput LaTeX environment and the \write18 TeX primitive to write and
run some shell and Python scripts. Those scripts extract code examples from other CUDA, Rust, Python
or other languages source files, run them and capture their output. I then define LaTeX commands
that use to script to insert the code samples and their output into the PDF documents. This
way, the code examples in the documents are always up to date and tested at build time, as part of
the PDF generation process.