repository·trunk·Indexed 12 days ago
https://github.com/SeleniumHQ/seleniumAn umbrella project providing infrastructure for the W3C WebDriver specification, enabling web browser automation across all major browsers via a language-neutral interface. It provides language bindings for .NET (Selenium.WebDriver), Java (selenium-java), and JavaScript (selenium-webdriver).
The Selenium .NET API is divided into two primary modules that provide different levels of functionality for browser automation:
The Selenium Python client driver supports several browsers and the Remote protocol:
To prevent confusion between supported Selenium APIs and unsupported BiDi implementation details, low-level access is provided via composition rather than as a member of the driver object.
Accessing a feature directly from the driver (e.g., driver.bidi) is considered an internal exposure and is not the intended way to interact with the protocol. Instead, you should compose the protocol module with the driver instance.
Example (Ruby):
driver.network.add_request_handler(...) — This is the correct way to use high-level, protocol-neutral features.BiDi::Protocol::Network.new(driver).add_intercept(...) — This is how you access the raw BiDi implementation via composition.driver.bidi.network.add_intercept(...) — This pattern is rejected because anything reachable directly off the driver is implicitly treated as a supported Selenium API.# supported — neutral, returns no BiDi type
driver.network.add_request_handler(...)
# internal — composed with the driver, unsupported
BiDi::Protocol::Network.new(driver).add_intercept(...)
# not allowed — internal exposed as a driver member
driver.bidi.network.add_intercept(...) The Python test suite provides several fixtures to simplify setup and teardown. Shared fixtures are located in conftest.py, while module-specific fixtures are defined within the test files themselves.
| Fixture | Description |
|---|---|
driver | WebDriver instance, auto-parametrized by browser |
pages | Load test pages: pages.load("page.html") or pages.url("page.html") |
webserver | Test HTTP server reference |
clean_driver | Fresh driver without parametrization |
clean_options | Fresh browser options instance |
For local scripts, a Java server is not required. However, to run Selenium tests remotely, you must run a Selenium Grid.
To use the Remote WebDriver in Python, refer to the official documentation for remote driver configuration.
The shared driver is reused across tests for efficiency. Use these annotations to control when the driver is restarted or destroyed. Annotations accept value (browser) and reason.
| Annotation | Behavior |
|---|---|
@NeedsFreshDriver | Restarts the driver before the test to ensure a clean browser state with default capabilities. |
@NoDriverBeforeTest | Driver is destroyed before the test. Use createNewDriver(capabilities) inside the test to create one with custom capabilities. |
@NoDriverAfterTest | Restarts the driver after the test because the test leaves the browser in a bad state. Supports failedOnly. |
Important Implementation Rules:
new ChromeDriver(). Always use WebDriverBuilder or createNewDriver(capabilities). The builder respects the browser target defined by Bazel.createNewDriver(capabilities) is called without an annotation, it closes the current driver and creates a new one.Tests are organized into unit and integration suites. All test files must follow the naming convention of ending in _spec.rb (e.g., driver_spec.rb).
Directory Structure:
rb/spec/unit/: Unit tests that do not require a browser.rb/spec/integration/: Integration tests involving browsers (Chrome, Firefox, Safari, Bidi) and spec_support helpers.Selenium requires a driver (e.g., chromedriver, edgedriver, geckodriver) to interface with the chosen browser.
Modern versions of Selenium automatically handle browser and driver installation using Selenium Manager. You generally do not need to manually download drivers or add them to your system PATH; Selenium Manager handles this when you instantiate a WebDriver.
If Selenium Manager does not meet your specific needs, you can still manually install and specify browsers and drivers.
When a WebDriver session is requested in a Kubernetes-enabled Grid:
Running state and detects early errors like ImagePullBackOff./status endpoint until it returns HTTP 200.To avoid re-downloading dependencies and rebuilding artifacts when using multiple Git worktrees, configure shared Bazel caches in your global ~/.bazelrc:
--disk_cache: Stores compiled action outputs.--repository_cache: Stores downloaded external dependencies.Note: These directories grow unbounded and should be pruned periodically. Keep them on the same filesystem as your checkouts to allow Bazel to use hardlinks.
To prevent the Bazel output base from leaking disk space when deleting worktrees, point the output base inside the worktree by adding this to the worktree's .bazelrc.local:
startup --output_base=.local/output-base.
Do not nest the output base inside the repo on Windows due to path-length limits. Instead, use startup --output_user_root=C:/tmp in selenium/.bazelrc.windows.local.
# Add to ~/.bazelrc
common --disk_cache=/path/to/your/home/.cache/bazel-disk
common --repository_cache=/path/to/your/home/.cache/bazel-repo
# Add to worktree/.bazelrc.local (macOS/Linux)
startup --output_base=.local/output-base