graphql-engine/server/tests-py/README.md

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

310 lines
15 KiB
Markdown
Raw Normal View History

# Python Integration Test Suite
This document describes the Python integration test suite. Please consult
the `server/CONTRIBUTING` document for general information on the overall
test setup and other testing suites.
This document describes running and writing tests, as well as some
information on how to update test dependencies.
2018-11-08 09:32:57 +03:00
## Running tests
Tests can be run using the `dev.sh` script or directly using `pytest`.
Please note that running the `BigQuery` tests requires a few manual steps.
### Running tests via dev.sh
The easiest way to run the test suite is to do:
```sh
scripts/dev.sh test --integration
```
NOTE: this only runs the tests for Postgres. If you want to run tests
for a different backend, use:
```sh
scripts/dev.sh test --integration --backend mssql
```
Available options are documented in `scripts/parse-pytest-backend`:
- postgres (default)
- bigquery (see section below)
- citus
- mssql
- mysql
#### Filtering tests
You can filter tests by using `-k <name>`. Note that `<name>` is case-
insensitive.
```sh
scripts/dev.sh test --integration --backend mssql -k MSSQL
```
Note that you can also use expressions here, for example:
```sh
scripts/dev.sh test --integration --backend mssql -k "MSSQL and not Permission"
```
See [pytest docs](https://docs.pytest.org/en/6.2.x/usage.html#specifying-tests-selecting-tests)
for more details.
#### Failures
If you want to stop after the first test failure you can pass `-x`:
```sh
scripts/dev.sh test --integration --backend mssql -k MSSQL -x
```
#### Verbosity
You can increase or decrease the log verbosity by adding `-v` or `-q`
to the command.
### Running tests directly
WARNING: running tests manually will force skipping of some tests. `dev.sh`
deals with setting up some environment variables which decide how and if
some of the tests are executed.
1. To run the Python tests, youll need to install the necessary Python dependencies first. It is recommended that you do this in a self-contained Python venv, which is supported by Python 3.3+ out of the box. To create one, run:
```
python3 -m venv .python-venv
```
(The second argument names a directory where the venv sandbox will be created; it can be anything you like, but `.python-venv` is `.gitignore`d.)
With the venv created, you can enter into it in your current shell session by running:
```
source .python-venv/bin/activate
```
(Source `.python-venv/bin/activate.fish` instead if you are using `fish` as your shell.)
2. Install the necessary Python dependencies into the sandbox:
```
pip3 install -r tests-py/requirements.txt
```
3. Install the dependencies for the Node server used by the remote schema tests:
```
(cd tests-py/remote_schemas/nodejs && npm ci)
```
4. Start an instance of `graphql-engine` for the test suite to use:
```
env EVENT_WEBHOOK_HEADER=MyEnvValue \
EVENT_WEBHOOK_HANDLER=http://localhost:5592 \
SCHEDULED_TRIGGERS_WEBHOOK_DOMAIN=http://127.0.0.1:5594 \
cabal new-run -- exe:graphql-engine \
--database-url='postgres://<user>:<password>@<host>:<port>/<dbname>' \
serve --stringify-numeric-types
```
Optionally, replace the `--database-url` parameter with `--metadata-database-url` to enable testing against multiple sources.
The environment variables are needed for a couple of tests, and the `--stringify-numeric-types` option is used to avoid the need to do floating-point comparisons.
run graphql tests on both http and websocket; add parallelism (close #1868) (#1921) Examples 1) ` pytest --hge-urls "http://127.0.0.1:8080" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" -vv ` 2) `pytest --hge-urls "http://127.0.0.1:8080" "http://127.0.0.1:8081" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" "postgresql://admin@127.0.0.1:5432/hge_tests2" -vv ` ### Solution and Design <!-- How is this issue solved/fixed? What is the design? --> <!-- It's better if we elaborate --> #### Reducing execution time of tests - The Schema setup and teardown, which were earlier done per test method, usually takes around 1 sec. - For mutations, the model has now been changed to only do schema setup and teardown once per test class. - A data setup and teardown will be done once per test instead (usually takes ~10ms). - For the test class to get this behaviour, one can can extend the class `DefaultTestMutations`. - The function `dir()` should be define which returns the location of the configuration folder. - Inside the configuration folder, there should be - Files `<conf_dir>/schema_setup.yaml` and `<conf_dir>/schema_teardown.yaml`, which has the metadata query executed during schema setup and teardown respectively - Files named `<conf_dir>/values_setup.yaml` and `<conf_dir>/values_teardown.yaml`. These files are executed to setup and remove data from the tables respectively. #### Running Graphql queries on both http and websockets - Each GraphQL query/mutation is run on the both HTTP and websocket protocols - Pytests test parameterisation is used to achieve this - The errors over websockets are slightly different from that on HTTP - The code takes care of converting the errors in HTTP to errors in websockets #### Parallel executation of tests. - The plugin pytest-xdist helps in running tests on parallel workers. - We are using this plugin to group tests by file and run on different workers. - Parallel test worker processes operate on separate postgres databases(and separate graphql-engines connected to these databases). Thus tests on one worker will not affect the tests on the other worker. - With two workers, this decreases execution times by half, as the tests on event triggers usually takes a long time, but does not consume much CPU.
2019-04-08 10:22:38 +03:00
5. Optionally, add more sources to test against:
If the tests include more sources (e.g., by using `-k MSSQL or MySQL`), then you can use the following commands to add sources to your running graphql instance:
```
# Add a Postgres source
curl "$METADATA_URL" \
--data-raw '{"type":"pg_add_source","args":{"name":"default","configuration":{"connection_info":{"database_url":"'"$POSTGRES_DB_URL"'","pool_settings":{}}}}}'
# Add a SQL Server source
curl "$METADATA_URL" \
--data-raw '{"type":"mssql_add_source","args":{"name":"mssql","configuration":{"connection_info":{"connection_string":"'"$MSSQL_DB_URL"'","pool_settings":{}}}}}'
# Optionally verify sources have been added
curl "$METADATA_URL" --data-raw '{"type":"export_metadata","args":{}}'
```
6. With the server running, run the test suite:
```
cd tests-py
pytest --hge-urls http://localhost:8080 \
--pg-urls 'postgres://<user>:<password>@<host>:<port>/<dbname>'
```
This will run all the tests, which can take a couple minutes (especially since some of the tests are slow). You can configure `pytest` to run only a subset of the tests; see [the `pytest` documentation](https://doc.pytest.org/en/latest/usage.html) for more details.
Some other useful points of note:
- It is recommended to use a separate Postgres database for testing, since the tests will drop and recreate the `hdb_catalog` schema, and they may fail if certain tables already exist. (Its also useful to be able to just drop and recreate the entire test database if it somehow gets into a bad state.)
- You can pass the `-v` or `-vv` options to `pytest` to enable more verbose output while running the tests and in test failures. You can also pass the `-l` option to display the current values of Python local variables in test failures.
- Tests can be run against a specific backend (defaulting to Postgres) with the `backend` flag, for example:
```
pytest --hge-urls http://localhost:8080 \
--pg-urls 'postgres://<user>:<password>@<host>:<port>/<dbname>'
--backend mssql -k TestGraphQLQueryBasicCommon
```
For more details, please consult `pytest --help`.
### Running BigQuery tests
Running integration tests against a BigQuery data source is a little more involved due to the necessary service account requirements:
```
HASURA_BIGQUERY_PROJECT_ID=# the project ID of the service account
HASURA_BIGQUERY_SERVICE_ACCOUNT_EMAIL=# eg. "<<SERVICE_ACCOUNT_NAME>>@<<PROJECT_NAME>>.iam.gserviceaccount.com"
HASURA_BIGQUERY_SERVICE_KEY=# the service account key
```
Before running the test suite:
1. Ensure you have access to a [Google Cloud Console service account](https://cloud.google.com/iam/docs/creating-managing-service-accounts#creating). Store the project ID and account email in `HASURA_BIGQUERY_PROJECT_ID` variable.
2. [Create and download a new service account key](https://cloud.google.com/iam/docs/creating-managing-service-account-keys). Store the contents of file in a `HASURA_BIGQUERY_SERVICE_KEY` variable.
```bash
export HASURA_BIGQUERY_SERVICE_KEY=$(cat /path/to/service/account)
```
3. [Login and activate the service account](https://cloud.google.com/sdk/gcloud/reference/auth/activate-service-account), if it is not already activated.
4. Verify the service account is accessible via the [BigQuery API](https://cloud.google.com/bigquery/docs/reference/rest):
1. Run the following command:
```bash
source scripts/verify-bigquery-creds.sh $HASURA_BIGQUERY_PROJECT_ID $HASURA_BIGQUERY_SERVICE_KEY $HASURA_BIGQUERY_SERVICE_ACCOUNT_EMAIL
```
If the query succeeds, the service account is setup correctly to run tests against BigQuery locally.
5. Finally, run the BigQuery test suite with `HASURA_BIGQUERY_SERVICE_KEY` and `HASURA_BIGQUERY_PROJECT_ID` environment variables set. For example:
```
scripts/dev.sh test --integration --backend bigquery -k TestGraphQLQueryBasicBigquery
```
_Note to Hasura team: a service account is already setup for internal use, please check the wiki for further details._
## Tests structure
run graphql tests on both http and websocket; add parallelism (close #1868) (#1921) Examples 1) ` pytest --hge-urls "http://127.0.0.1:8080" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" -vv ` 2) `pytest --hge-urls "http://127.0.0.1:8080" "http://127.0.0.1:8081" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" "postgresql://admin@127.0.0.1:5432/hge_tests2" -vv ` ### Solution and Design <!-- How is this issue solved/fixed? What is the design? --> <!-- It's better if we elaborate --> #### Reducing execution time of tests - The Schema setup and teardown, which were earlier done per test method, usually takes around 1 sec. - For mutations, the model has now been changed to only do schema setup and teardown once per test class. - A data setup and teardown will be done once per test instead (usually takes ~10ms). - For the test class to get this behaviour, one can can extend the class `DefaultTestMutations`. - The function `dir()` should be define which returns the location of the configuration folder. - Inside the configuration folder, there should be - Files `<conf_dir>/schema_setup.yaml` and `<conf_dir>/schema_teardown.yaml`, which has the metadata query executed during schema setup and teardown respectively - Files named `<conf_dir>/values_setup.yaml` and `<conf_dir>/values_teardown.yaml`. These files are executed to setup and remove data from the tables respectively. #### Running Graphql queries on both http and websockets - Each GraphQL query/mutation is run on the both HTTP and websocket protocols - Pytests test parameterisation is used to achieve this - The errors over websockets are slightly different from that on HTTP - The code takes care of converting the errors in HTTP to errors in websockets #### Parallel executation of tests. - The plugin pytest-xdist helps in running tests on parallel workers. - We are using this plugin to group tests by file and run on different workers. - Parallel test worker processes operate on separate postgres databases(and separate graphql-engines connected to these databases). Thus tests on one worker will not affect the tests on the other worker. - With two workers, this decreases execution times by half, as the tests on event triggers usually takes a long time, but does not consume much CPU.
2019-04-08 10:22:38 +03:00
- Tests are grouped as test classes in test modules (names starting with `test_`)
- The configuration files (if needed) for the tests in a class are usually kept in one folder.
- The folder name is usually either the `dir` variable or the `dir()` function
run graphql tests on both http and websocket; add parallelism (close #1868) (#1921) Examples 1) ` pytest --hge-urls "http://127.0.0.1:8080" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" -vv ` 2) `pytest --hge-urls "http://127.0.0.1:8080" "http://127.0.0.1:8081" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" "postgresql://admin@127.0.0.1:5432/hge_tests2" -vv ` ### Solution and Design <!-- How is this issue solved/fixed? What is the design? --> <!-- It's better if we elaborate --> #### Reducing execution time of tests - The Schema setup and teardown, which were earlier done per test method, usually takes around 1 sec. - For mutations, the model has now been changed to only do schema setup and teardown once per test class. - A data setup and teardown will be done once per test instead (usually takes ~10ms). - For the test class to get this behaviour, one can can extend the class `DefaultTestMutations`. - The function `dir()` should be define which returns the location of the configuration folder. - Inside the configuration folder, there should be - Files `<conf_dir>/schema_setup.yaml` and `<conf_dir>/schema_teardown.yaml`, which has the metadata query executed during schema setup and teardown respectively - Files named `<conf_dir>/values_setup.yaml` and `<conf_dir>/values_teardown.yaml`. These files are executed to setup and remove data from the tables respectively. #### Running Graphql queries on both http and websockets - Each GraphQL query/mutation is run on the both HTTP and websocket protocols - Pytests test parameterisation is used to achieve this - The errors over websockets are slightly different from that on HTTP - The code takes care of converting the errors in HTTP to errors in websockets #### Parallel executation of tests. - The plugin pytest-xdist helps in running tests on parallel workers. - We are using this plugin to group tests by file and run on different workers. - Parallel test worker processes operate on separate postgres databases(and separate graphql-engines connected to these databases). Thus tests on one worker will not affect the tests on the other worker. - With two workers, this decreases execution times by half, as the tests on event triggers usually takes a long time, but does not consume much CPU.
2019-04-08 10:22:38 +03:00
- Some tests (like in `test_graphql_queries.py`) requires a setup and teardown per class.
run graphql tests on both http and websocket; add parallelism (close #1868) (#1921) Examples 1) ` pytest --hge-urls "http://127.0.0.1:8080" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" -vv ` 2) `pytest --hge-urls "http://127.0.0.1:8080" "http://127.0.0.1:8081" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" "postgresql://admin@127.0.0.1:5432/hge_tests2" -vv ` ### Solution and Design <!-- How is this issue solved/fixed? What is the design? --> <!-- It's better if we elaborate --> #### Reducing execution time of tests - The Schema setup and teardown, which were earlier done per test method, usually takes around 1 sec. - For mutations, the model has now been changed to only do schema setup and teardown once per test class. - A data setup and teardown will be done once per test instead (usually takes ~10ms). - For the test class to get this behaviour, one can can extend the class `DefaultTestMutations`. - The function `dir()` should be define which returns the location of the configuration folder. - Inside the configuration folder, there should be - Files `<conf_dir>/schema_setup.yaml` and `<conf_dir>/schema_teardown.yaml`, which has the metadata query executed during schema setup and teardown respectively - Files named `<conf_dir>/values_setup.yaml` and `<conf_dir>/values_teardown.yaml`. These files are executed to setup and remove data from the tables respectively. #### Running Graphql queries on both http and websockets - Each GraphQL query/mutation is run on the both HTTP and websocket protocols - Pytests test parameterisation is used to achieve this - The errors over websockets are slightly different from that on HTTP - The code takes care of converting the errors in HTTP to errors in websockets #### Parallel executation of tests. - The plugin pytest-xdist helps in running tests on parallel workers. - We are using this plugin to group tests by file and run on different workers. - Parallel test worker processes operate on separate postgres databases(and separate graphql-engines connected to these databases). Thus tests on one worker will not affect the tests on the other worker. - With two workers, this decreases execution times by half, as the tests on event triggers usually takes a long time, but does not consume much CPU.
2019-04-08 10:22:38 +03:00
- Here we are extending the `DefaultTestSelectQueries` class.
- This class defines a fixture which will run the configurations in `setup.yaml` and `teardown.yaml` once per class
- Extending test class should define a function name `dir()`, which returns the configuration folder
run graphql tests on both http and websocket; add parallelism (close #1868) (#1921) Examples 1) ` pytest --hge-urls "http://127.0.0.1:8080" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" -vv ` 2) `pytest --hge-urls "http://127.0.0.1:8080" "http://127.0.0.1:8081" --pg-urls "postgresql://admin@127.0.0.1:5432/hge_tests" "postgresql://admin@127.0.0.1:5432/hge_tests2" -vv ` ### Solution and Design <!-- How is this issue solved/fixed? What is the design? --> <!-- It's better if we elaborate --> #### Reducing execution time of tests - The Schema setup and teardown, which were earlier done per test method, usually takes around 1 sec. - For mutations, the model has now been changed to only do schema setup and teardown once per test class. - A data setup and teardown will be done once per test instead (usually takes ~10ms). - For the test class to get this behaviour, one can can extend the class `DefaultTestMutations`. - The function `dir()` should be define which returns the location of the configuration folder. - Inside the configuration folder, there should be - Files `<conf_dir>/schema_setup.yaml` and `<conf_dir>/schema_teardown.yaml`, which has the metadata query executed during schema setup and teardown respectively - Files named `<conf_dir>/values_setup.yaml` and `<conf_dir>/values_teardown.yaml`. These files are executed to setup and remove data from the tables respectively. #### Running Graphql queries on both http and websockets - Each GraphQL query/mutation is run on the both HTTP and websocket protocols - Pytests test parameterisation is used to achieve this - The errors over websockets are slightly different from that on HTTP - The code takes care of converting the errors in HTTP to errors in websockets #### Parallel executation of tests. - The plugin pytest-xdist helps in running tests on parallel workers. - We are using this plugin to group tests by file and run on different workers. - Parallel test worker processes operate on separate postgres databases(and separate graphql-engines connected to these databases). Thus tests on one worker will not affect the tests on the other worker. - With two workers, this decreases execution times by half, as the tests on event triggers usually takes a long time, but does not consume much CPU.
2019-04-08 10:22:38 +03:00
- For mutation tests (like in `test_graphql_mutations.py`)
- We need a `schema_setup` and `schema_teardown` per class
- And `values_setup` and `values_teardown` per test
- Doing schema setup and teardown per test is expensive.
- We are extending the `DefaultTestMutations` class for this.
- This class defines a fixture which will run the configuration in `setup.yaml` and `teardown.yaml` once per class.
- Another fixture defined in this class runs the configuration in `values_setup.yaml` and `values_teardown.yaml` once per class.
## Writing python tests
1. Check whether the test you intend to write already exists in the test suite, so that there will be no
duplicate tests or the existing test will just need to be modified.
2. All the tests use setup and teardown, the setup step is used to initialize the graphql-engine
and the database in a certain state after which the tests should be run. After the tests are run,
the state needs to be cleared, which should be done in the teardown step. The setup and teardown
is localised for every python test class.
See `TestCreateAndDelete` in [test_events.py](test_events.py)
for reference.
3. The setup and teardown can be configured to run before and after every test in a test class
or run before and after running all the tests in a class. Depending on the use case, there
are different fixtures like `per_class_tests_db_state`,`per_method_tests_db_state` defined in the [conftest.py](conftest.py) file.
4. Sometimes, it's required to run the graphql-engine with in a different configuration only
for a particular set of tests. In this case, these tests should be run only when the graphql-engine
is run with the said configuration and should be skipped in other graphql-engine configurations. This
can be done by accepting a new command-line flag from the `pytest` command and depending on the value or
presence of the flag, the tests should be run accordingly. After adding this kind of a test, a new section
needs to be added in the [test-server.sh](../../oss-.circleci/test-server.sh). This new section's name should also
be added in the `server-test-names.txt` file, otherwise the test will not be run in the CI.
For example,
The tests in the [test_remote_schema_permissions.py](test_remote_schema_permissions.py)
are only to be run when the remote schema permissions are enabled in the graphql-engine and when
it's not set, these tests should be skipped. Now, to run these tests we parse a command line option
from pytest called (`--enable-remote-schema-permissions`) and the presence of this flag means that
we need to run these tests. When the tests are run with this command line option, it's assumed that
the server has enabled remote schema permissions.
### Adding test support for a new backend
The current workflow for supporting a new backend in integration tests is as follows:
1. Add functions to launch and cleanup a server for the new backend. [Example](https://github.com/hasura/graphql-engine/blob/5df8419a4f0efb3032e07130e5d5f33e75e53b66/scripts/containers/postgres).
2. Augment `dev.sh` to support the new backend. [Example](https://github.com/hasura/graphql-engine/blob/5df8419a4f0efb3032e07130e5d5f33e75e53b66/scripts/dev.sh#L208-L214).
3. Connect the GraphQL Engine to the database you've just launched. [Example](https://github.com/hasura/graphql-engine/blob/5df8419a4f0efb3032e07130e5d5f33e75e53b66/scripts/data-sources-util.sh).
4. Add setup and teardown files:
1. `setup_<backend>`: for `v1/query` or metadata queries such as `<backend>_track_table`. [Example](https://github.com/hasura/graphql-engine/blob/5df8419a4f0efb3032e07130e5d5f33e75e53b66/server/tests-py/queries/graphql_query/basic/setup_mssql.yaml).
2. `schema_setup_<backend>`: for `v2/query` queries such as `<backend>_run_sql`. [Example](https://github.com/hasura/graphql-engine/blob/5df8419a4f0efb3032e07130e5d5f33e75e53b66/server/tests-py/queries/graphql_query/basic/schema_setup_mssql.yaml).
3. `teardown_<backend>` and `cleardb_<backend>`
4. **Important:** filename suffixes _**should be the same**_ as the value thats being passed to `—backend`; that's how the files are looked up.
5. Specify a `backend` parameter for [the `per_backend_test_class` and `per_backend_test_function` fixtures](https://github.com/hasura/graphql-engine/blob/5df8419a4f0efb3032e07130e5d5f33e75e53b66/server/tests-py/conftest.py#L311-L333), parameterised by backend. [Example](https://github.com/hasura/graphql-engine/blob/5df8419a4f0efb3032e07130e5d5f33e75e53b66/server/tests-py/test_graphql_queries.py#L123).
Note: When teardown is not disabled (via `skip_teardown`(\*) , in which case, this phase is skipped entirely), `teardown.yaml` always runs before `schema_teardown.yaml`, even if the tests fail. See `setup_and_teardown` in `server/tests-py/conftest.py` for the full source code/logic.
(\*): See `setup_and_teardown_v1q` and `setup_and_teardown_v2q` in `conftest.py` for more details.
This means, for example, that if `teardown.yaml` untracks a table, and `schema_teardown.yaml` runs raw SQL to drop the table, both would succeed (assuming the table is tracked/exists).
**Test suite naming convention**
The current convention is to indicate the backend(s) tests can be run against in the class name. For example:
- `TestGraphQLQueryBasicMSSQL` for tests that can only be run against a SQL Server backend
- `TestGraphQLQueryBasicCommon` for tests that can be run against more than one backend
- If a test class doesn't have a suffix specifying the backend, nor does its name end in `Common`, then it is likely a test written pre-v2.0 that
can only be run on Postgres
This naming convention enables easier test filtering with [pytest command line flags](https://docs.pytest.org/en/6.2.x/usage.html#specifying-tests-selecting-tests).
The backend-specific and common test suites are disjoint; for example, run `pytest --integration -k "Common or MySQL" --backend mysql` to run all MySQL tests.
Note that `--backend` does not interact with the selection of tests. You will generally have to combine `--backend` with `-k`.
## Updating Python requirements
The packages/requirements are documented in two files:
- `server/tests-py/requirements-top-level.txt`
- `server/tests-py/requirements.txt`
The `server/tests-py/requirements-top-level.txt` file is the main file. It
contains the direct dependencies along with version requirements we know
we should be careful about.
The `server/tests-py/requirements.txt` file is the _lock_ file. It holds
version numbers for all direct and transitive dependencies. This file
can be re-generated by:
1. alter `server/tests-py/requirements-top-level.txt`
2. remove `server/tests-py/requirements.txt`
3. run `dev.sh test --integration`
4. update `DEVSH_VERSION` in `scripts/dev.sh` to force reinstall
these dependencies
Steps 3 can be done manually:
```sh
pip3 install -r requirements-top-level.txt
pip3 freeze > requirements.txt
```