* add -image locator strategy and related logic
* use appium-support jimp helper
* handle find multiple elements by image
* update staleness check variable name
* ensure that a settings object always exists so we can rely on global settings always existing too
* only fix template size if user requests it in settings
* add note about implicit wait error handling
* only automatically update image element position if user turns on an appropriate setting
* add setting to control whether we check for image element staleness
* add setting to determine the image element tap strategy (w3c or mjsonwp)
* prefer single quotes
* add setting to turn off screenshot resizing if not desired
* add docstrings (except for cache class which we may not keep)
* address additional review comments
* add better jsdoc arg info
* use lru-cache npm package instead of rolling our own
* use new versino of appium-support
* make sure the lru cache is checking element size correctly
* give better names and tighten up logic around image el tap strategies
* add some unit tests for image elements
* ensure DeviceSettings has a default no-op update method
* check that driver implements action strategies and fall back / throw as appropriate
* finish out image element unit tests
* have basedriver set noop onSettingsUpdate rather than doing it in DeviceSettings
* move route avoidance detection from protocol to driver
* handle proxying actions to downstream drivers during image tap
* tighten up logic for deciding whether to proxy based on presence of image element
* add find by image specs and fix issues that arose in the process
* clean up regexes
* don't try to be smart and proxy actions for drivers; let them set it up themselves to support image elements
* change default match threshold based on experimentation
* export which settings are handled by basedriver so downstream drivers can utilize this information
* make sure to include a duration for the initial pointerMove action
* add docstring to proxyRouteIsAvoided
* Properly handle MJSONWP responses with non-zero status and return code < 400
* code 500 looks more appopriate
* Optimize downstream protocol detection
* Update the unit test
* Update errors handling inside proxyReqRes
* Improve the condition
* This test worked fine in base driver but these tests also run in fake driver and don't work beause `app ` is a required capability
* Added `app=fake` to caps
* Before, reset was creating a new session and using the processed caps
* The problem with is, if it was W3C it would create a MJSONWP
* Now it checks the protocol before resetting and passes in the appropriate args to createSession depending on the protocol type
* Before it was throwing immediately which meant we weren't parsing the protocol and weren't showing the proper error responses
* If proxied W3C error doesn't have an error, make it `Unknown Error`
* Some problems with parsing `{protocol: <protocol>, value: <value>}`
* `(!_.isNil(driverRes.value)` was incorrect check because `value` can be undefined or null; replaced it with `_.has(driverRes, 'value')` to check that the `value` property was set
* ```
if (driverRes.value) {
driverRes = driverRes.value;
}
```
was also incorrect because `driverRes.value` may be false-y
* Moved logic into it's own function and then added tests to cover the problem cases
* If`executeCommand` returns an object that has a property 'protocol' and either a 'value' property or an 'error' property, respond with the provided protocol
* If the object is of this form, than it came from another driver that knows what protocol is being used so we re-assign the protocol to whatever was reported
* Added to delay in the 'unexpected exits' test because I was getting a race condition
* Add common websocket handlers
* tune the default prefix
* Add integration test
* Tune the test
* Fix the functional test
* Tune socket import
* Address review comments
* Tune some expectations
* Previously, we were only checking the HTTP payload to see what protocol should be
* This is inadequate, in the case of drivers that don't proxy W3C Capabilities yet
* The fix now is to set the protocol in `createSession` to see if the W3C caps were proxied through, if they're not, set protocol to MJSONWP
* Added unit test and one e2e test to confirm that this works
* Convert JSONWP proxy errors
* Previously, if a proxied response has a status code >= 0 but http status code 200, the error is thrown but it's an unknown error because the 'getResponseForW3CError' function doesn't translate the error
* Now, it translates the JSONWP error using logic we already have that accomplishes this
* Added E2E test that tests a proxy that returns 200 but status 7 to check that it throws the correct ElementNotFoundError
* Added unit test to handle the new case in 'getResponseForW3CError'
* Do proxying for W3C responses
* If a W3C Response is provided by a proxy, map to the correct error
* Checks if `value` is an object and `value.error` and maps it based on that (`value.error` is the W3C error code)
* Broke off `errorFromCode` into `errorFromMJSONWPStatusCode` and `errorFromW3CJsonCode`
* Rename '.error' property to correct W3C property
* The error property in the ProtocolError objects should match up with https://www.w3.org/TR/webdriver/#handling-errors under the 'JSON Error Code' column
* Added tests for all of these
* Make ProxyRequestError work for W3C
* Added an arg called `.w3c` that passes in the proxied W3C error
* Fixed the `getActualError` function that looks for the .w3c variable and parses error from the code
* Rename 'mjsonwp' to 'protocol'
* Renamed folder from 'mjsonwp' to 'protocol' in `lib/` and `test/`
* Renamed 'mjsonwp.js' to 'protocol.js'
* Renamed any import statements that use 'mjsonwp'
* _Possible breaking change_: MJSONWireProtocol class exposed in index was renamed to just Protocol.... checked Appium repo and found no libraries that use MJSONWireProtocol
* Lint fix
* Fixed HTTPResponseBody for W3C error
* Refactored `getResponseForW3CError` into errors.js file
* Changed method so that it returns `{value: {stack: '...', message: '...', error: '...'}}`
* Message is the same as the JSONWP error message
* Error is a static message indicating the type of error
* 'stack' is the callstack provided in the error
* Added W3C 'error' messages to all of the errors
* Make 'getResponseForW3CError' handle BadParametersError
* Refactor 'getResponseForJsonwpError' to errors.js
* If server encounters a response that has the wrong key name (ELEMENT when it should be 'element-6066-11e4-a52e-4f735466cecf'), rename all instances of that key
* Added a helper method (with tests) that recursively finds instances of a 'key' and renames it
* Added e2e test that mocks `findElements`, calls `/elements` endpoint and verifies that the payload was correctly renamed
* Previously, it was going through alwaysMatch and all of the firstMatch caps and individually logging when any of them had unprefixed, non-standard caps
* Now it does them all at once
* Made a helper function that takes a caps object ({alwaysMatch, firstMatch}) and finds all of the non-prefixed caps in all of them
* Calls that function and logs warning
* Moved 'ProxyRequestError' handling outside of the MJSONWP block.
* When it's a proxy error, Appium translates the error (based on statusCode) to it's proper Error
* W3C weren't doing this when ProxyRequestError was being used
* Now W3C translates proxy errors
* Added tests in `mjsonwp-e2e-specs` that test JWProxy
* Test successful 200 proxies
* Tested unsuccessful responses to see that a JSONWP response with status = 6 would return the correct 'NoSuchDriverError' with statusCode = 404
* If 'capabilities' and 'desiredCapabilities' were provided, and 'capabilities' is NOT a plain object, fall back to 'desiredCapabilities'
* Added more logging to the W3C vs. MJSONWP determination logic
* Rename `createSession` args to be more descriptive
* Added basic W3C unit tests for above fix
* Make malformed caps return 400 error
* Add InvalidArgumentTypeError to errors
* Made 'NotYetImplementedError' the 405 (Method Not Allowed) status code
* Made errors in 'createSession' return InvalidArgumentTypeError
* Fix incorrect W3C responses
* Add w3c status for NotImplemted and NotYetImplemented
* Changed `getResponseForW3CStatusError` so that throws an unknown error if the exception isn't a w3c exception
* Checks for property `w3cStatus` to confirm if it's a proper protocol-blessed exception
* Moved logic that reassigned errors into MJSONWP blocks; W3C shouldn't reassign errors because they all have proper status codes
* Added endpoint tests that check that you can create sessions in W3C and that subsequent calls to the session return the correct responses
* fix timeouts for w3c
* revert lock file
* add validation for previous timeouts
* implement custom validator for W3C
* improve if condition
* define isW3CProtocol
* rewrite validation a bit
* apply util.hasValue and string protocol
* add tests
* refactor timeouts for mjsonwp and w3c
* use const
* remove previous timeouts
* use undefined
* rename to timeouts again
* wrap some commands
* remove ...args
* define each command
Mykola Mokhnach