Common guide #
This section has articles related to all submodules (frontend, backend and exporter) such as: code style hints, architecture decisions, etc...
Configuration #
Both in the backend, the frontend and the exporter subsystems, there are an
app.config
namespace that defines the global configuration variables,
their specs and the default values.
All variables have a conservative default, meaning that you can set up a Penpot instance without changing any configuration, and it will be reasonably safe and useful.
In backend and exporter, to change the runtime values you need to set them in
the process environment, following the rule that an environment variable in the
form PENPOT_<VARIABLE_NAME_IN_UPPERCASE>
correspond to a configuration
variable named variable-name-in-lowercase
. Example:
(env)
PENPOT_ASSETS_STORAGE_BACKEND=assets-s3
(config)
assets-storage-backend :assets-s3
In frontend, the main resources/public/index.html
file includes (if it
exists) a file named js/config.js
, where you can set configuration values
as javascript global variables. The file is not created by default, so if
you need it you must create it blank, and set the variables you want, in
the form penpot<VariableNameInCamelCase>
:
(js/config.js)
var penpotPublicURI = "https://penpot.example.com";
(config)
public-uri "https://penpot.example.com"
On premise instances #
If you use the official Penpot docker images, as explained in the Getting Started section, there is a config.env file that sets the configuration environment variables. It's the same file for backend, exporter and frontend.
For this last one, there is a script
nginx-entrypoint.sh
that reads the environment and generates the js/config.js
when the container
is started. This way all configuration is made in the single config.env
file.
Dev environment #
If you use the developer docker images, the docker-compose.yaml directly sets the environment variables more appropriate for backend and exporter development.
Additionally, the backend start script and repl script set some more variables.
The frontend uses only the defaults.
If you want to change any variable for your local environment, you can change
docker-compose.yaml
and shut down and start again the container. Or you can
modify the start script or directly set the environment variable in your
session, and restart backend or exporter processes.
For frontend, you can manually create resources/public/js/config.js
(it's
ignored in git) and define your settings there. Then, just reload the page.
System logging #
In app.common.logging we have a general system logging utility, that may be used throughout all our code to generate execution traces, mainly for debugging.
You can add a trace anywhere, specifying the log level (trace
, debug
,
info
, warn
, error
) and any number of key-values:
(ns app.main.data.workspace.libraries-helpers
(:require [app.common.logging :as log]))
(log/set-level! :warn)
...
(defn generate-detach-instance
[changes container shape-id]
(log/debug :msg "Detach instance"
:shape-id shape-id
:container (:id container))
...)
The current namespace is tracked within the log message, and you can configure
at runtime, by namespace, the log level (by default :warn
). Any trace below
this level will be ignored.
Some keys have a special meaning:
:msg
is the main trace message.::log/raw
outputs the value without any processing or prettifying.::log/context
append metadata to the trace (not printed, it's to be processed by other tools).::log/cause
(only in backend) attach a java exception object that will be printed in a readable way with the stack trace.::log/async
(only in backend) if set to false, makes the log processing synchronous. If true (the default), it's executed in a separate thread.:js/<key>
(only in frontend) if you prefix the key with thejs/
namespace, the value will be printed as a javascript interactively inspectionable object.:err
(only in frontend) attach a javascript exception object, and it will be printed in a readable way with the stack trace.
backend #
The logging utility uses a different library for Clojure and Clojurescript. In the first case we use log4j2 to have much flexibility.
The configuration is made in log4j2.xml
file. The Logger used for this is named "app" (there are other loggers for
other subsystems). The default configuration just outputs all traces of level
debug
or higher to the console standard output.
There is a different log4j2-devenv
for the development environment. This one outputs traces of level trace
or
higher to a file, and debug
or higher to a zmq
queue, that may be
subscribed for other parts of the application for further processing.
The ouput for a trace in logs/main.log
uses the format
[<date time>] : <level> <namespace> - <key1=val1> <key2=val2> ...
Example:
[2022-04-27 06:59:08.820] T app.rpc - action="register", name="update-file"
The zmq
queue is not used in the default on premise or devenv setups, but there
are a number of handlers you can use in custom instances to save errors in the
database, or send them to a Sentry or similar
service, for example.
frontend and exporter #
In the Clojurescript subservices, we use goog.log library. This is much simpler, and basically outputs the traces to the console standard output (the devtools in the browser or the console in the nodejs exporter).
In the browser, we have an utility debug function that enables you to change the logging level of any namespace (or of the whole app) in a live environment:
debug.set_logging("namespace", "level")
Assertions #
Penpot source code has this types of assertions:
assert #
Just using the clojure builtin assert
macro.
Example:
(assert (number? 3) "optional message")
This asserts are only executed in development mode. In production environment all asserts like this will be ignored by runtime.
spec/assert #
Using the app.common.spec/assert
macro.
This macro is based in cojure.spec.alpha/assert
macro, and it's
also ignored in a production environment.
The Penpot variant doesn't have any runtime checks to know if asserts are disabled. Instead, the assert calls are completely removed by the compiler/runtime, thus generating simpler and faster code in production builds.
Example:
(require '[clojure.spec.alpha :as s]
'[app.common.spec :as us])
(s/def ::number number?)
(us/assert ::number 3)
spec/verify #
An assertion type that is always executed.
Example:
(require '[app.common.spec :as us])
(us/verify ::number 3)
This macro enables you to have assertions on production code, that generate runtime exceptions when failed (make sure you handle them appropriately).
Unit tests #
We expect all Penpot code (either in frontend, backend or common subsystems) to have unit tests, i.e. the ones that test a single unit of code, in isolation from other blocks. Currently we are quite far from that objective, but we are working to improve this.
Running tests with kaocha #
Unit tests are executed inside the development environment.
We can use kaocha test runner, and
we have prepared, for convenience, some aliases in deps.edn
files. To run
them, just go to backend
, frontend
or common
and execute:
# To run all tests once
clojure -M:dev:test
# To run all tests and keep watching for changes
clojure -M:dev:test --watch
# To run a single tests module
clojure -M:dev:test --focus common-tests.logic.comp-sync-test
# To run a single test
clojure -M:dev:test --focus common-tests.logic.comp-sync-test/test-sync-when-changing-attribute
Watch mode runs all tests when some file changes, except if some tests failed previously. In this case it only runs the failed tests. When they pass, then runs all of them again.
You can also mark tests in the code by adding metadata:
;; To skip a test, for example when is not working or too slow
(deftest ^:kaocha/skip bad-test
(is (= 2 1)))
;; To skip it but warn you during test run, so you don't forget it
(deftest ^:kaocha/pending bad-test
(is (= 2 1)))
Please refer to the kaocha manual for how to define custom metadata and other ways of selecting tests.
NOTE: in frontend
we still can't use kaocha to run the tests. We are on
it, but for now we use shadow-cljs with package.json
scripts:
yarn run test
yarn run test:watch
Test output #
The default kaocha reporter outputs a summary for the test run. There is a pair
of brackets [ ]
for each suite, a pair of parentheses ( )
for each test,
and a dot .
for each assertion t/is
inside tests.
penpot@c261c95d4623:~/penpot/common$ clojure -M:dev:test
[(...)(............................................................
.............................)(....................................
..)(..........)(.................................)(.)(.............
.......................................................)(..........
.....)(......)(.)(......)(.........................................
..............................................)(............)]
190 tests, 3434 assertions, 0 failures.
All standard output from the tests is captured and hidden, except if some test fails. In this case, the output for the failing test is shown in a box:
FAIL in sample-test/stdout-fail-test (sample_test.clj:10)
Expected:
:same
Actual:
-:same +:not-same
╭───── Test output ───────────────────────────────────────────────────────
│ Can you see this?
╰─────────────────────────────────────────────────────────────────────────
2 tests, 2 assertions, 1 failures.
You can bypass the capture with the command line:
clojure -M:dev:test --no-capture-output
Or for some specific output:
(ns sample-test
(:require [clojure.test :refer :all]
[kaocha.plugin.capture-output :as capture]))
(deftest stdout-pass-test
(capture/bypass
(println "This message should be displayed"))
(is (= :same :same)))
Running tests in the REPL #
An alternative way of running tests is to do it from inside the REPL you can use in the backend and common apps in the development environment.
We have a helper function (run-tests)
that refreshes the environment (to avoid
having stale tests)
and runs all tests or a selection. It is defined in backend/dev/user.clj
and
common/dev/user.clj
, so it's available without importing anything.
First start a REPL:
~/penpot/backend$ scripts/repl
And then:
;; To run all tests
(run-tests)
;; To run all tests in one namespace
(run-tests 'some.namespace)
;; To run a single test
(run-tests 'some.namespace/some-test)
;; To run all tests in one or several namespaces,
;; selected by a regular expression
(run-tests #"^backend-tests.rpc.*")
Writing unit tests #
We write tests using the standard Clojure test API. You can find a guide to writing unit tests at Practicalli Clojure, that we follow as much as possible.
Sample files helpers #
An important issue when writing tests in Penpot is to have files with the specific configurations we need to test. For this, we have defined a namespace of helpers to easily create files and its elements with sample data.
To make handling of uuids more convenient, those functions have a uuid
registry. Whenever you create an object, you may give a :label
, and the id of
the object will be stored in the registry associated with this label, so you
can easily recover it later.
You have functions to create files, pages and shapes, to connect them and specify their attributes, having all of them default values if not set.
Files also store in metadata the current page, so you can control in what
page the add-
and get-
functions will operate.
(ns common-tests.sample-helpers-test
(:require
[app.common.test-helpers.files :as thf]
[app.common.test-helpers.ids-map :as thi]
[app.common.test-helpers.shapes :as ths]
[clojure.test :as t]))
(t/deftest test-create-file
(let [;; Create a file with one page
f1 (thf/sample-file :file1)
;; Same but define the label of the page, to retrieve it later
f2 (thf/sample-file :file2 :page-label :page1)
;; Set the :name attribute of the created file
f3 (thf/sample-file :file3 :name "testing file")
;; Create an isolated page
p2 (thf/sample-page :page2 :name "testing page")
;; Create a second page and add to the file
f4 (-> (thf/sample-file :file4 :page-label :page3)
(thf/add-sample-page :page4 :name "other testing page"))
;; Create an isolated shape
p2 (thf/sample-shape :shape1 :type :rect :name "testing shape")
;; Add a couple of shapes to a previous file, in different pages
f5 (-> f4
(ths/add-sample-shape :shape2)
(thf/switch-to-page :page4)
(ths/add-sample-shape :shape3 :name "other testing shape"
:width 100))
;; Retrieve created shapes
s1 (ths/get-shape f4 :shape1)
s2 (ths/get-shape f5 :shape2 :page-label :page3)
s3 (ths/get-shape f5 :shape3)]
;; Check some values
(t/is (= (:name f1) "Test file"))
(t/is (= (:name f3) "testing file"))
(t/is (= (:id f2) (thi/id :file2)))
(t/is (= (:id (thf/current-page f2)) (thi/id :page1)))
(t/is (= (:id s1) (thi/id :shape1)))
(t/is (= (:name s1) "Rectangle"))
(t/is (= (:name s3) "testing shape"))
(t/is (= (:width s3) 100))
(t/is (= (:width (:selrect s3)) 100))))
Also there are functions to make some transformations, like creating a component, instantiating it or swapping a copy.
(ns app.common-tests.sample-components-test
(:require
[app.common.test-helpers.components :as thc]
[app.common.test-helpers.files :as thf]
[app.common.test-helpers.shapes :as ths]))
(t/deftest test-create-component
(let [;; Create a file with one component
f1 (-> (thf/sample-file :file1)
(ths/add-sample-shape :frame1 :type :frame)
(ths/add-sample-shape :rect1 :type :rect
:parent-label :frame1)
(thc/make-component :component1 :frame1))]))
Finally, there are composition helpers, to build typical structures with a
single line of code. And the files module has some functions to display the
contents of a file, in a way similar to debug/dump-tree
but showing labels
instead of ids:
(ns app.common-tests.sample-compositions-test
(:require
[app.common.test-helpers.compositions :as tho]
[app.common.test-helpers.files :as thf]))
(t/deftest test-create-composition
(let [f1 (-> (thf/sample-file :file1)
(tho/add-simple-component-with-copy :component1
:main-root
:main-child
:copy-root))]
(ctf/dump-file f1 :show-refs? true)))
;; {:main-root} [:name Frame1] # [Component :component1]
;; :main-child [:name Rect1]
;;
;; :copy-root [:name Frame1] #--> [Component :component1] :main-root
;; <no-label> [:name Rect1] ---> :main-child
You can see more examples of usage by looking at the existing unit tests.