This project is now archived. All code is still available, but is unlikely of much use (you can read about some of the design issues in the blog below and in the doc folder).
News: Pippin now has a blog!
Pippin is a database inspired by distributed version control systems (notably git). Unlike git it is designed to store thousands to millions (or more) small objects in only a few dozen files. Unlike most databases, it is designed to allow independent distributed usage with eventual (potentially user-assisted) synchronisation.
Pippin is an object database in that complex objects (any supporting serialisation) are stored directly, not a relational database using predefined numeric and string types. Pippin is serverless, designed to run in a single process, or even from multiple processes simultaneously using the same files on disk (with eventual synchronisation).
ACID-compliant transactions are possible (with a single DB host and within a single partition), depending on how merges are handled; merge functionality is highly configurable.
The full history of transactions/commits is recorded via a combination of snapshot files and commit logs. Snapshot files never change and logs are kept small, making for simple backups. Old snapshots and logs can simply be deleted if the history is no longer required.
Pippin is designed to be scalable via partitioning, achieved using user-defined classifiers. This functionality, along with filtering (i.e. simple search), is planned, but not yet available.
For more, see the documentation in src/lib.rs or take a look at the examples.
Pippin is 'alpha' status.
Partition-oriented usage (i.e. a single 'partition') should have all the basic features there and is ready for testing, but the API may change. Perhaps the biggest caveat is that every commit is written to a new file due to not yet working out how to safely extend files.
Repository-oriented usage is not yet ready.
What should work:
- persistance of data within a single 'partition' via snapshots
- storing changes via commit logs
- reconstruction of state from snapshot + logs
- auto-detecting latest state(s)
- merging of multiple latest states (may require user-interaction)
- checksumming & detecting corrupt stuff
- recovery of some data when files are missing (though this needs more work)
- file formats are mosly final except that headers will get extra data and object diffs
What is planned:
- tracking mutliple partitions in a distributed manner via file headers
- user-specified classifiers
- (possibly) indexes of some kind
- reclassification of objects as necessary
- partially-automated division of "large" partitions via classifiers
- object diffs (current commits include a full copy of all changed entries)
- log file extension (currently a new file is used per commit to avoid data loss)
The doc directory contains some file-format documentation and various notes planning Pippin's development.
Tickets were originally stored in files. Several "tags" are still in use; where applicable these are mentioned in tickets and can be used to find relevant bits of code. All of these can be found with grep:
egrep -IR "#00[0-9]{2}" doc/ src/
Some self-contained examples can be found in the examples and tests
directories:
examples/hello.rs — minimal example
examples/pippincmd.rs — tool to read/write DB entries as plain text
app_tests/examples/sequences.rs — test program generating random DB entries
tests/partition-ops.rs — external test suite for partition operations
app_tests/tests/seq_create_small.rs — create a small random repo as a test
More examples and tests can be found in the applications directory. These
make use of an extra library including some common code.
Pippin uses Cargo. A few example commands:
cargo test
cargo build --release
cargo run --example pippincmd -- -h
cargo help run
cargo doc && open target/doc/pippin/index.html
Generated binaries can be found in the target directory.
Pippin is licenced under the Mozilla Public License, version 2.0. A copy of this licence can be found in the LICENSE-MPL2.txt file or obtained at http://mozilla.org/MPL/2.0/ .
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the MPL-2.0 license, shall be licensed as above, without any additional terms or conditions.