jest-csl

jest-csl

This is a library to make testing Citation Style Language definitions easier using Jest. It includes:

• A testing-specific abstraction over the unwieldy citeproc-js
• A mechanism to fetch and cache locales and the Juris-M style-modules (i.e. jurisdiction-specific CSL-M modules), and let the processor access them.
• A schema for writing CSL specs in YAML
• A simple function to execute one Jest test for each test in a spec, for a particular CSL-JSON reference library, for a particular style
• A CLI and an API to to collect results for further processing (like turning it into HTML).

How to test your CSL

Prerequisites

• You should be familiar with using a shell. You should be able to open up a terminal, and use ls, cd, mkdir and friends.
• You need to have Node.js installed. Run node -v to check it's version 10 or later.
• You should install yarn with npm install -g yarn, or mentally replace yarn in the rest of this document with npm.
• You must have Git installed.

Set up your project

This step is fairly flexible, but the following will work and is a good place to start.

my-style-repo
├── src
│   ├── style.csl
│   └── juris-au-style.csl (jurisdiction override, if you are using CSL-M)
├── test
│   ├── corpus.json
│   ├── style.test.js
│   └── style.spec.yaml
├── package.json
├── .gitignore
└── jest.config.js

Name your items according to whatever your style is called, like chicago-author-date.csl instead of style.csl.

package.json

Add a package.json with the following contents:

{
  "scripts": {
    "test": "jest --watchAll --noStackTrace"
  }
}

Then, run yarn add -D jest jest-csl. You will get the latest version of jest and this library (jest-csl).

test/corpus.json

This file should not be built by hand, it is too tedious. Instead:

1. Install Zotero Better Bibtex in your choice of Zotero or Juris-M.
2. Create a library for testing your style, optionally add some items to it
3. Export your library with Right-Click > Export Library ...; select Better CSL JSON and tick the Keep updated box.
4. When selecting the destination, choose /path/to/your-style-repo/test/corpus.json.

test/style.spec.yaml

This is where you write your tests. See below.

test/style.test.js: a test configuration

Insert the following contents, modifying STYLE and optionally un-commenting or deleting jurisdictionDirs as required for your own style:

module.exports = {
  csl: "./src/STYLE.csl",
  // use jurisdictionDirs if you are using CSL-M and want to include overrides
  // jurisdictionDirs: ["./src"],
  libraries: ["./test/corpus.json"],
  suites: ["./test/STYLE.spec.yaml"]
};

const { jestCSL } = require('jest-csl');
jestCSL(module.exports);

jest.config.js

This file lets jest-csl check that the locales are all cached before running.

module.exports = {
  globalSetup: "jest-csl/setup"
};

It won't update the cache. That would be too slow to do on every test run. See 'Updating the cache' below.

.gitignore

You're probably going to want to put all of this in Git at some point, so add these contents:

node_modules

Writing tests

Note: This test suite was developed for a footnote-based style. If you're building a different kind of style (i.e. trying to test bibliography output), we might need to add a new kind of test, so please file an issue.

Specs are written in YAML. If you haven't written YAML before, read this primer.

The YAML spec files are structured somewhat like a Jest test suite. Here's an example:

- describe: "Name of a feature of this CSL to test (often a kind of document)"
  tests:

    - it: "should render a basic book citation"
      single: { id: "doe2001" }
      # you can use > to join multiple lines with spaces if a line gets long
      expect: "Doe, <i>Miscellaneous Writings</i>, 2001."

    - it: "should include both volume and issue in a journal article"
      single: { id: "doe2003" }
      # you can use > to join multiple lines with spaces if a line gets long
      # <i></i> is italics, for other formats just run the test first
      expect: >
        John Doe, 'A Journal Article Written in My Diary' (2003) 89(3) 
        <i>John's Diary</i> 124.

Basic single-item tests

Clusters and Sequences

For more complex combined citations, use 'sequence' to test the in-texts/footnotes generated for a sequence of clusters of cites.

- describe: "..."
  tests:
    ...
- describe: "Subsequent references"
  tests:
    - it: "should render plain ibids for the same locator"
      sequence:
        - [ { id: "doe2001", locator: "5", label: "page" }
          , { id: "doe2001", locator: "5", label: "page", prefix: "see also ", suffix: " etc" } ]
        - [{ id: "doe2001", locator: "5", label: "page" }]
      expect:
        - Doe, <i>Miscellaneous Writings</i>, 2001, p. 5; see also <i>ibid</i> etc.
        - <i>Ibid</i>.

Abbreviations

To test the abbreviations found in the Abbreviation Filter:

- describe: "a"
  tests:
    - it: "should x"
      ...
      abbreviations:
        - hereinafter:
            doe2001: Misc
        - jurisdiction: us
          container-title:
            Coolest Tribunal Of The Land: CTL

Without a jurisdiction specified, it is default.

See this command's output for a list of allowed categories.

node -e 'var c=require('citeproc');console.log(Object.keys(new c.AbbreviationSegments()))'

Unit-Testing Specific Macros

You can add macro: MacroName to a test, and for that test only, your style's <layout> tag will have its contents replaced with <text macro="MacroName" />, and any prefix and suffix is stripped (delimiter is kept). This is convenient when you are building a style from the ground up and can't write full-item tests yet.

<macro name="MacroName">
  <text variable="locator" />
</macro>
<citation ...>
  <layout prefix="(" suffix=")" delimiter="; " ...>
    <text value="lots of messy code" />
    <choose>
      ...
    </choose>
  </layout>
</citation>

- describe: "MacroName only"
  tests:
    - it: "should render a locator"
      macro: MacroName
      single: { id: "citekey", locator: "a-locator" }
      expect: "a-locator"

Combining test suites

Sometimes you will have sets of tests that you want to pass for two different styles, and then extended suites that only apply to one style. This might encompass two variations on one style, or just the plain CSL vs CSL-M versions of your style. It would make sense to have more than one test configuration (.test.js file) to exercise more than one combination.

You can pass multiple suites in your test configuration file. They will be merged according to the order; later styles with the same describe > it combinations will override previous ones.

# core.spec.yaml
- describe: "Unit"
  tests:
    - it: "should a"
      ...
    - it: "should b"
      ...

# extended.spec.yaml
- describe: "Unit"
  tests:
    - it: "should b"
      single: "override"
      expect: "override"
    - it: "should c"
      single: "new"
      expect: "new"

Setting

{
  // ...
  suites: ["./test/core.spec.yaml", "./test/extended.spec.yaml"]
}

Results in all three tests being run, with "should a" preserved, "should b" overridden and "should c" added.

Further splitting of test suites

For convenience, you can split a large test suite into multiple files, and combine them all with a glob. Use a level of directories or a naming scheme to separate groups that should be strictly before or after one another.

{
  // ...
  suites: ["./test/core/*.yaml", "./test/extended/*.yaml"]
}

The same works for jurisdictionDirs and libraries.

Skipping or isolating tests

You can add a mode to a test with the value skip or only or known. If it's skip, it will be skipped; if >= 1 test has only, every test that isn't labelled only will be skipped. known means it will be skipped in Jest, but will be included in the results output (for documenting failures but not polluting your test runner).

Also, any test that does not have an expect key will be skipped as a test stub, so you can write lots of unwritten it keys as a to-do list.

- describe: "Unit"
  tests:
    - it: "should b"
      mode: skip
      ...

Running your tests

yarn test

Follow the onscreen instructions to interact with the Jest environment (or quit).

Updating the cache

yarn jest-csl update

This will attempt to update your cached locales and style-modules.

Collecting test results with the CLI

Before, you defined a test configuration that would export a plain-object description of itself if it wasn't running in Jest. You can use that now.

1. Define an output key on the configuration object, with a string path to a JSON file to be written.
2. Run jest-csl results on your configuration.

// ./test/my-configuration.test.js
module.exports = {
    ...,
    output: './results.json'
}

yarn jest-csl results --includeLibrary ./test/my-configuration.test.js
jq . results.json

{
    "library": {
        "citekey": { ... },
        ...
    },
    "units": [
        {
            "describe": "...",
            "tests": [...]
        }
    ]
}

Each test object has these keys added:

• type indicating what kind of test it was:
  • single
  • sequence
  • doc for documentation only (i.e. it had a doc key but no actual test)
  • stub otherwise
• result key of the same shape as expect (only if the test had an expect)

Test Metadata

Two keys, doc and meta, are recognised as metadata on both describe units and test cases. They may have any content, and will be included in the output.

Collecting test results via JavaScript

const { cslTestResults } = require('jest-csl');
const config = require('./test/my-configuration.test');
let { engine, units, library, citeIds } = cslTestResults(config);
console.log(units);

This generates a JS array of each of the test units, with each test transformed as described in the CLI. This is useful for generating documentation or making a custom view of the results, a bit like a jest reporter but without losing the test case information and metadata.

engine is a wrapper around CSL.Engine pre-loaded with your libraries. engine.retrieveItem('citeKey') may be useful to you.