This project isn't concerned with creating a complete digital representation of a Magic card. Rather, it aims to JSON-ise the rules text in order to enable better querying. For example, consider searching for hexproof as an ability. This finds cards like Geist of Saint Traft, which has this ability unconditionally, but also cards like Dragonlord Ojutai, which has it conditionally, and Bristling Hydra, which acquires it through an activated ability. There's no (sane) string-based query to get around this.
Therefore, the aim is to build a JSON construct which can embody these concepts in a queryable fashion. This is theoretically doable manually, but doing so for ~16,000 cards would be an insane task. Slightly less insane is to treat the rules text as a programming language. This involves a grammar parser like Antlr4. The .g4 files describe the grammar of Magic-ese (Magic English? Magese?) and parse it into a syntax tree, which the (forthcoming) python code will be able to traverse and subsequently generate a JSON representation.
First, download Antlr v4 from the site. This should result in a jar file named something like 'antlr-4.6-complete.jar'. Theoretically, a text editor is all that's needed for editing .g4 files, but why make things more difficult?
Antlrworks2 is an IDE built specifically for .g4 files, but to test grammar files you have to load them up in a test rig, which slows development of the grammar.
IntelliJ Idea (the community version) is a better solution. It can handle most languages, and with the Antlr4 plugin, it generates live syntax trees on text that you can enter in a sidebox, allowing for instant testing of grammars. In Idea, File->Settings and then Plugins gets you to a plugin list. Search for Antlr v4 (click 'search in repositories' if you don't see a result) and install it. Once that's done, open the filigree-text folder. Then to test the syntax, open the mtg.g4 file and right click on the evaluate rule, and select Test Rule evaluate. Then use the input text box to test the grammar, or select a file.
Now you can edit grammar files conveniently. However, we need to use the syntax trees the grammar generates, and while the default choice is Java, I'm going full Python. In a terminal, go to the folder containing the grammar files and run this command:
java -cp antlr-4.6-complete.jar org.antlr.v4.Tool mtg.g4 -Dlanguage=Python3
where mtg.g4 will be the name of the top-level grammar file (it imports other ones). (You can leave out the -Dlanguage=Python3 option to generate default Java classes.) This generates Python classes (a lexer, a listener and a parser) which will be used in a future Python file which will read in a (future) corpus file and pass each card into the syntax tree generator.