Store

Store is a configurable virtual file system inside liquer. Store is designed to be able to deal with states. One notable extension of the Store compared to a regular file system is the ability to store (and work with) the metadata, which is essential for dealing with data in liquer.

Store is basically a key/value store mapping a path to a sequence of bytes. By itself, Store does not define (or care) about serialization of the data. This differenciates it from Cache.

Even though the interface to Cache and Store is intentionaly very similar, these two mechanisms are different:

• Cache keeps State (i.e. data object with metadata). Cache deals with objects and stores State perhaps in a non-serialized form (e.g. MemoryStore).
• Store keeps resources - i.e. arbitrary binary data (bytes) complemented with metadata.

One purpose of a Store is to provide an option to serve files into the pipeline. The pipeline may start with a resource path followed by a sequence of actions.

Cache

By default there is no cache - i.e. the queries are always re-evaluated. There are several cache implementations available in liquer.cache. They are configured by set_cache function, for example

set_cache(FileCache("cache"))

configures a cache that will store all the (chache-able) results of queries in a dictionary cache. Cache should be configured before the queries are evaluated - and before state variables are set.

Currently there are three cache implementations: NoCache is a trivial do-nothing cache, FileCache stores data in files, MemoryCache caches the object in the memory.

Custom cache can be created by defining a cache interface, see above mentioned classes. Cache will typically use query as a key and utilize the mechanism of serializing data into a bytes sequence (defined in liquer.state_types), thus implementing a cache based either on a key-value store or blob-storage in SQL databases should be fairly straightforward (and probably quite similar to FileCache).

Command may optionally decide not to cache its output. This may be useful when command produces volatile data, e.g. time. In such a case command (operating on a state) can disable cache by state.with_caching(False).