Skip to content

Latest commit

 

History

History
366 lines (257 loc) · 14.3 KB

dev-practices.md

File metadata and controls

366 lines (257 loc) · 14.3 KB

Description

This page describes development practices for this codebase.

Linting

Most of our linters require babashka. Before running them, please install babashka. To invoke all the linters in this section, run

bb dev:lint

Clojure code

To lint:

clojure -M:clj-kondo --parallel --lint src --cache false

We lint our Clojure(Script) code with https://github.com/clj-kondo/clj-kondo/. If you need to configure specific linters, see this documentation. Where possible, a global linting configuration is used and namespace specific configuration is avoided.

There are outstanding linting items that are currently ignored to allow linting the rest of the codebase in CI. These outstanding linting items should be addressed at some point:

  • Comments starting with TODO:lint
  • Code marked with #_:clj-kondo/ignore require a good understanding of the context to address as they usually involve something with a side effect or require changing multiple fns up the call stack.

Unused vars

We use https://github.com/borkdude/carve to detect unused vars in our codebase.

To run this linter:

bb lint:carve

By default, the script runs in CI mode which prints unused vars if they are found. The script can be run in an interactive mode which prompts for keeping (ignoring) an unused var or removing it. Run this mode with:

bb lint:carve '{:interactive true}'

When a var is ignored, it is added to .carve/ignore. Please add a comment for why a var is ignored to help others understand why it's unused.

Large vars

Large vars have a lot of complexity and make it hard for the team to maintain and understand them. To run this linter:

bb lint:large-vars

To configure the linter, see the [:tasks/config :large-vars] path of bb.edn.

Document namespaces

Documentation helps teams share their knowledge and enables more individuals to contribute to the codebase. Documenting our namespaces is a good first step to improving our documentation. To run this linter:

bb lint:ns-docstrings

To skip documenting a ns, use the common ^:no-doc metadata flag.

Datalog linting

We use datascript's datalog to power our modeling and querying layer. Since datalog is concise, it is easy to write something invalid. To avoid typos and other preventable mistakes, we lint our queries and rules. Our queries are linted through clj-kondo and datalog-parser. clj-kondo will error if it detects an invalid query.

Translations

We use tongue, a simple and effective library, for translations. We have a couple bb tasks for working with translations under lang: e.g. bb lang:list. See the translator guide for usage.

One useful task for reviewers (us) and contributors alike, is bb lang:validate-translations which catches common mistakes). When reviewing translations here are some things to keep in mind:

  • Punctuation and delimiting characters (e.g. :, :, ?) should be part of the translatable string. Those characters and their position may vary depending on the language.
  • Translations usually return strings but they can return hiccup vectors with a fn translation. Hiccup vectors are needed when word order matters for a translation and formatting is involved. See this 3 word Turkish example.
  • Translations can be anonymous fns with arguments for interpolating strings. Fns should be simple and only include the following fns: str, when, if and =.

Spell Checker

We use typos to spell check our source code.

To install it locally and use it:

$ brew install typos-cli
# Catch any errors
$ typos
# Fix errors
$ typos -w

To configure it e.g. for dealing with false positives, see typos.toml.

Testing

We have unit, performance and end to end tests.

End to End Tests

Even though we have a nightly release channel, it's hard for testing users (thanks to the brave users!) to notice all issues in a limited time, as Logseq is covering so many features. The only solution is automatic end-to-end tests - adding tests for GUI software is always painful but necessary. See https://github.com/logseq/logseq/pulls?q=E2E for e2e test examples.

To run end to end tests

yarn electron-watch
# in another shell
yarn e2e-test # or npx playwright test

If e2e failed after first running:

There's a traceAll() helper function to enable playwright trace file dump for specific test files logseq#8332

If e2e tests fail in the file, they can be debugged by examining a trace dump with the playwright trace viewer.

Locally this will get dumped into e2e-dump/.

On CI the trace file will be under Artifacts at the bottom of a run page e.g. https://github.com/logseq/logseq/actions/runs/3574600322.

Unit Testing

Our unit tests use the shadow-cljs test-runner. To run them:

yarn test

By convention, a namespace's tests are found at a corresponding namespace of the same name with an added -test suffix. For example, tests for frontend.db.model are found in frontend.db.model-test.

There are a couple different ways to run tests:

There a couple types of tests and they can overlap with each other:

Focus Tests

Tests can be selectively run on the commandline using our own test runner which provides the same test selection options as cognitect-labs/test runner. For this workflow:

  1. Run clj -M:test watch test in one shell
  2. Focus tests:
  3. Add ^:focus metadata flags to tests e.g. (deftest ^:focus test-name ...).
  4. In another shell, run node static/tests.js -i focus to only run those tests. To run all tests except those tests run node static/tests.js -e focus.
  5. Or focus namespaces: Using the regex option -r, run tests for frontend.util.page-property-test with node static/tests.js -r page-property.

Multiple options can be specified to AND selections. For example, to run all frontend.util.page-property-test tests except for the focused one: node static/tests.js -r page-property -e focus

For help on more options, run node static/tests.js -h.

Autorun Tests

To run tests automatically on file save, run clojure -M:test watch test --config-merge '{:autorun true}'. Specific namespace(s) can be auto run with the :ns-regexp option e.g. clojure -M:test watch test --config-merge '{:autorun true :ns-regexp "frontend.util.page-property-test"}'.

REPL tests

Most unit tests e.g. ones that are browser compatible and don't require node libraries, can be run from the REPL. To do so:

  • Start a REPL for your editor. See here for an example.
  • Load a test namespace.
  • Run (cljs.test/run-tests) to run tests for the current test namespace.

Database tests

To write a test that uses a datascript db:

  • Be sure your test ns has test fixtures from test-helper ns to create and destroy test databases after each test.
  • The easiest way to set up test data is to use test-helper/load-test-files.
  • For the repo argument that most fns take, pass it test-helper/test-db

Performance tests

To write a performance test:

  • Use frontend.util/with-time-number to get the time in ms.

  • Example:

    (are [x timeout] (>= timeout (:time (util/with-time-number (block/normalize-block x true))))
        ... )

For examples of these tests, see frontend.db.query-dsl-test and frontend.db.model-test.

Async Tests

Async unit testing is well supported in ClojureScript. https://clojurescript.org/tools/testing#async-testing is a good guide for how to do this. We have a couple of test helpers that make testing async easier:

  • frontend.test.helper/deftest-async - deftest for async tests that ensures uncaught exceptions don't abruptly end the test suite. If you don't use this macro for async tests, you are expected to handle unexpected failures in your test
  • frontend.test.helper/with-reset - A version of with-redefs that works for async contexts

Accessibility

Please refer to our accessibility guidelines.

Logging

For logging, we use https://github.com/lambdaisland/glogi. When in development, be sure to have enabled custom formatters in the desktop app and browser. Without this enabled, most of the log messages aren't readable.

Data validation and generation

We use malli and spec data validation, fn validation (and generation someday). malli has the advantage that its schema is data and can be used for additional purposes.

Reusable malli schemas should go under src/main/frontend/schema/ and be compatible with clojure and clojurescript. See frontend.schema.handler.plugin-config for an example.

Reusable specs should go under src/main/frontend/spec/ and be compatible with clojure and clojurescript. See frontend.spec.storage for an example.

By following these conventions, these should also be usable by babashka. This is helpful as it allows for third party tools to be written with logseq's data model.

Optionally Validating Functions

We use malli for optionally validating fns a.k.a instrumenting fns. Function validation is enabled in dev mode. To add typing for a fn, just add it to a var's metadata per this example. We also have clj-kondo type annotations derived from these fn schemas. To re-generate them after new schemas have been added, update the namespaces in gen-malli-kondo-config.core and then run bb dev:gen-malli-kondo-config. To learn more about fn instrumentation, see this page.

Auto-formatting

Currently the codebase is not formatted/indented consistently. We loosely follow https://github.com/bbatsov/clojure-style-guide. cljfmt is a common formatter used for Clojure, analogous to Prettier for other languages. You can do so easily with the Calva extension in VSCode: It will (mostly) indent your code correctly as you type, and you can move your cursor to the start of the line(s) you've written and press Tab to auto-indent all Clojure forms nested under the one starting on the current line.

Naming

We strive to use explicit names that are self explanatory so that our codebase is readable and maintainable. Sometimes we use abbreviations for frequently occurring concepts. Some common abbreviations:

  • rpath - Relative path e.g. logseq/config.edn
  • fpath - Full path e.g. /full/path/to/logseq/config.edn

Development Tools

Babashka tasks

There are a number of bb tasks under dev: for developers. Some useful ones to point out:

  • dev:validate-repo-config-edn - Validate a repo config.edn

    bb dev:validate-repo-config-edn src/resources/templates/config.edn
  • dev:publishing - Build a publishing app for a given graph dir. If the publishing frontend is out of date, it builds that first which takes time. Subsequent runs are quick.

    # One time setup
    $ cd scripts && yarn install && cd -
    
    # Build a release publishing app
    $ bb dev:publishing /path/to/graph-dir tmp/publish
    
    # OR build a dev publishing app that watches frontend changes
    $ bb dev:publishing /path/to/graph-dir tmp/publish --dev
    
    # View the publishing app in a browser
    $ python3 -m http.server 8080 -d tmp/publish &; open http://localhost:8080
    
    # Rebuild the publishing backend for dev/release.
    # Handy when making backend changes in deps/publishing or
    # to test a different graph
    $ bb dev:publishing-backend /path/graph-dir tmp/publish
    

There are also some tasks under nbb: which are useful for inspecting database changes in realtime. See these docs for more info.

Dev Commands

In the app, you can enable Dev commands under Settings > Advanced > Developer mode. Then search for commands starting with (Dev). Commands include inspectors for block/page data and AST.

Desktop Developer Tools

Since the desktop app is built with Electron, a full set of Chromium developer tools is available under the menu View > Toggle Developer Tools. Handy tools include a JS console and HTML inspector.

Security Practices

  • Our builds should not include unverified, third-party resources as this opens up the app to possibly harmful injections. If a third-party resource is included, it should be verified against an official distributor. Use logseq#9712 as an example to include a third party resource and not the examples under resources/js/.

FAQ

If dev app launch failed after electron upgrade:

yarn
yarn watch

In another window:

cd static
yarn
cd ..
yarn dev-electron-app

and kill all electron process Then a normal start happens via yarn dev-electron-app