Organize tests

test/rendering/readme.md

@@ -0,0 +1,40 @@
+# Rendering Tests
+
+The rendering tests use [Puppeteer](https://github.com/GoogleChrome/puppeteer) to take a screenshot of a map and compare it to an expected screenshot.  Each directory in the `cases` directory includes a single test case with the following files:
+
+ * `main.js` - the script that sets up the map and calls the magic `render()` function
+ * `expected.png` - the expected screenshot (this can be generated with the `--fix` flag)
+ * `actual.png` - the screenshot generated when running the tests (ignored by git)
+ * `pass` - a generated marker file that represents the last time the test passed (ignored by git)
+
+## Running the tests
+
+To run all the rendering tests:
+
+```bash
+node test/rendering/test.js
+```
+
+To run a single rendering test case:
+
+```bash
+node test/rendering/test.js --match your-test-case-name
+```
+
+If you want to leave the test server running (and the test browser open) after running a test, use the `--interactive` option.
+
+```bash
+node test/rendering/test.js --match your-test-case-name --interactive
+```
+
+## Creating a new test
+
+To create a new test case, add a directory under `cases` and add a `main.js` to it (copied from one of the other cases).  Then to generate the `expected.png` screenshot, run the test script with the `--fix` flag:
+
+```bash
+node test/rendering/test.js --match your-test-case-name --fix
+```
+
+Note that your `main.js` needs to call the magic `render()` function after setting up the map.  This triggers a snapshot.  The `render()` function can be called with an options object that includes a `tolerance` property.  The tolerance is related to the ratio of mismatched pixels to the total number of pixels in the actual screenshot.
+
+After creating a new test case, commit your `main.js` and `expected.png` files.