Testing files for Algorand SDKs

The files in this repository are used for testing the different Algorand SDK implementations. By writing the tests once and sharing them amongst the SDKs we are able to increase the coverage of our tests, and avoid rewriting similar tests over and over again. In addition to test cases, we have a standard test environment which is managed by docker.

To define tests we use cucumber, and feature files written with gherkin syntax. Each SDK is responsible for finding a framework which can use these files. There are implementations for many popular programming languages.

We have different feature files for unit and integration tests. The unit tests should be run as a normal part of development to quickly identify bugs and regressions. Integration tests on the other hand take much longer to run and require a special test environment. The test environment is made up of multiple services and managed with docker compose.

These reside in the unit features directory

These reside in the integration features directory

However, a few are not fully supported:

Full featured Algorand SDKs have 6 major components. Depending on the compatibility level, certain components may be missing. The components include:

1. REST Clients
2. Transaction Utilities
3. Encoding Utilities
4. Crypto Utilities
5. TEAL Utilities
6. Testing

The most basic functionality includes the REST clients for communicating with algod and indexer. These interfaces are defined by OpenAPI specifications:

• algod v1 / indexer v1 (generated at build time at daemon/algod/api/swagger.json)
• kmd v1 (generated at build time at daemon/kmd/api/swagger.json)
• algod v2
• indexer v2

One of the basic features of an Algorand SDK is the ability to construct all types of Algorand transactions. This includes simple transactions of all types and the tooling to configure things like leases and atomic transfers (group transactions)

In order to ensure transactions are compact and can hash consistently, there are some special encoding requirements. The SDKs must provide utilities to work with these encodings. Algorand uses MessagePack as a compact binary-encoded JSON alternative, and fields with default values are excluded from the encoded object. Additionally to ensure consistent hashes, the fields must be alphebatized.

All things related to crypto to make it easier for developers to work with the blockchain. This includes standard things like ED25519 signing, up through Algorand specific LogicSig and MultiSig utilities. There are also some convenience methods for converting Mnemonics.

Everything related to working with TEAL. This includes some utilities for parsing and validating compiled TEAL programs.

Each SDK has a number of unit tests specific to that particular SDK. The details of SDK-specific unit tests are up to the developers discretion. There are also a large number of cucumber integration tests stored in this repository which cover various unit-style tests and many integration tests. To assist with working in this environment each SDK must provide tooling to download and install the cucumber files, and a Dockerfile which configures an environment suitable for building the SDK and running the tests, and 3 makefile targets: make unit, make integration, and make docker-test. The rest of this document relates to details about the Cucumber test.

Tests consist of two things -- the feature files defined in this repository and some code snippets that map the text in the feature files to specific functions. The implementation process will vary by programming language and isn't covered here, refer to the relevant documentation for setting up a new SDK.

We use tags, and a simple directory structure, to organize our feature files. All cucumber implementations should allow specifying one or more tags to include, or exclude, when running tests.

All unit tests should be tagged with @unit so that unit tests can be run together during development for quick regression tests. For example, to run unit tests with java a tag filter is provided as follows:

~$ mvn test -Dcucumber.filter.tags="@unit"

This command will vary by cucumber implementation, the specific framework documentation should be referenced for details.

When adding a new test to an existing feature file, or a new feature file, a new tag should be created which describes that test. For example, the templates feature file has a corresponding @templates tag. By adding a new tag for each feature we are able to add new tests to this repository without breaking the SDKs.

In order for this to work, each SDK maintains a whitelist of tags which have been implemented.

If a new feature file is created, the tag would go at the top of the file. If a new scenario is added the tag would go right above the scenario.

If possible, please run a formatter on the file modified. There are several, including one built into VSCode Cucuber/Gherkin plugin.

The code snippets (or step definitions) live in the SDKs. Each SDK has a script which is able to clone this repository, and copy the tests into the correct locations.

When a test fails, the cucumber libraries we use print the code snippets which should be included in the SDK test code. The code snippets are empty functions which should be implemented according to the tests requirements. In many cases some state needs to be modified and stored outside of the functions in order to implement the test. Exactly how this state is managed is up to the developer. Refer to the cucumber documentation for tips about managing state. There may be better documentation in the specific cucumber language library you're using.

The SDKs come with a Makefile to coordinate running the cucumber test suites. There are 3 main targets:

• unit: runs all of the short unit tests.
• integration: runs all integration tests.
• harness: downloads this repo and calls up.sh to stand up a sandbox ready for running tests
• docker-test: installs feature file dependencies, starts the test environment, and runs the SDK tests in a docker container.

At a high level, the docker-test target is required to:

1. clone algorand-sdk-testing
2. copy supported feature files from the features directory into the SDK
3. build and start the test environment by calling ./scripts/up.sh which clones sandbox and stands it up
4. run all cucumber tests against the sandbox containers

This will vary by SDK. By calling up.sh the environment is available to the integration tests, and tests can be run locally with an IDE or debugger. This is often significantly faster than waiting for the entire test suite to run.

Some of the tests are stateful and will require restarting the environment before re-running the test.

Once the test environment is running you can use make unit and make integration to run tests.

Algorand's sandbox is used to manage several containers which work together to provide the test environment. This includes algod, kmd, indexer and a postgres database. The services run on specific ports with specific API tokens. Refer to .env and to sandbox'es docker-compose.yml for how these are configured.

up.sh is used to bring up the test environment. Not surprisingly, down.sh brings it all down.

When starting the environment, we default to using go-algorand's nightly build. If you're interested in running tests against a specific branch of go-algorand, you should set TYPE="source" in .env and set ALGOD_URL, and either ALGOD_BRANCH or ALGOD_SHA appropriately.

indexer and even the sandbox itself can be configured similarly through .env.