Caching
Tach makes use of a 'computation cache' to speed up certain tasks, such as testing.
When Tach finds cached results for a given task, you will see the terminal output enclosed in:
Caching is done at the command-level, meaning a single invocation of tach test can only ever result in a single cache hit or a single cache miss; individual tests are not cached separately.
Cache content
The computation cache contains the output from stdout and stderr from a previous task invocation.
This is done to enable 'replaying' cached tasks so that their output can be reused seamlessly.
Determining cache hits
Tach uses several pieces of information to determine cache hits. These include:
- Python interpreter version (
major.minor.micro) - All Python file contents beneath the source root
- Declared versions of 3rd party dependencies in project requirements (
requirements.txtorpyproject.toml) - File contents of explicitly configured file dependencies
- Explicitly configured environment variable values
When all of these pieces of information match a previous cache entry, the cached results are printed directly to the terminal.
Cache storage
The computation cache currently exists within the .tach directory in your project root. The directory is managed by Tach, which means that your cached results are stored on-disk on each machine where tasks are run.
We are currently working on a remote cache backend, which will allow multiple developers and CI environments to share a centralized cache to maximize the hit rate. If you are interested in this functionality, tell us on Discord, through a GitHub issue, or reach out via email: evan@gauge.sh; caelean@gauge.sh
Disabling the cache
The computation cache is enabled by default for commands such as tach test. For any command which utilizes the cache, it can be disabled using the --disable-cache CLI flag. This will prevent all access to the cache and run the underlying command unconditionally.