python-tkvdb 0.2.1

Cython wrapper for tkvdb radix trie key-value database

python-tkvdb

Python-tkvdb is a Cython wrapper for tkvdb trie key-value database. Python 3 is required.

Code isn't well tested in production environments.

Installation

This is a typical python/cython package that uses setuptools build system.

From PyPi

The most simple way of installing is using pip:

pip install python-tkvdb

Considering that package is using Cython, C compiler may be required for building if suitable wheel for current platform isn't found.

Downloading

Original tkvdb sources are included as submodule, so when installing from git please use git clone --recursive for cloning, or init submodules after with git submodule update --init --recursive. Custom sources may also be provided in tkvdb/ subdirectory before build.

Source archives that were made with python setup.py sdist contain generated C-code without .pyx files as Cython official documentation recommends. Cython header files (.pxd) still provided though.

Package also can be distributed as python wheels. With wheels no additional post-installation actions are required.

Initialization and usage of virtualenv or alternatives aren't properly described in this manual, so use them by you own discretion.

Building

Project uses Cython to generate C extension for Python. To build package, you need to have C compiler and build tools.

For installation from the source archive, Cython isn't required, but git versions require it. Source directory also includes pyproject.toml (PEP 518), so if your build tool uses it, Cython would be installed anyway.

To make common tasks easier, project contains simple Makefile that may be used instad of pip/python commands. It isn't a requirement, so pip install . also works. For additional reference, look into Makefile.

Install example with virtualenv:

cd python-tkvdb/
python -m venv env
source ./env/bin/activate
make

Both Makefile and setup.py uses USE_CYTHON env variable (int, 0/1) to determine if Cython (cythonize) would be started before extension building. Cython needs to be installed in local environment for this. Default value is 1 for make.

Makefile also has PYTHON env var that allows overriding python interpreter path. Default value is just python.

Example usage:

USE_CYTHON=1 PYTHON=./env/bin/python make build

Make commands:

• build -- just build extension with python setup.py build_ext.
• install -- run pip install. Extension would be built if needed.
• no arguments (just make) -- alias for install.
• dist -- create wheel and sdist archive.
• test -- run unit tests
• clean -- remove generated code, compiled objects and distribution archives.
• uninstall -- remove previously installed package (through pip)

After installing module tkvdb must be importable in the Python environment.

Usage

Original tkvdb uses pretty specific terminology for some actions (like transaction), so it is recommended to first consult with original documentation anyway. Thread-safety notes and some caveats are also described in the original README file.

Python-tkvdb provides pythonic-style wrapper for most parts of the original library.

Most object have destructors that call free internally and do memory management on garbage collection, but don't forget that this is wrapper for C library.

Some code examples may also be found in tests code.

Modules

Project is splitted into multiple python modules:

• ctkvdb -- Cython wrapper with C definitions from tkvdb.h.
• tkvdb.db -- database object and initialization. Also imported in __init__.py (i.e. main tkvdb module).
• tkvdb.transaction -- transaction (actually main input/output. interface). Wrapper around tkvdb_tr.
• tkvdb.cursor -- transaction cursors for iteration. Wrapper around tkvdb_cursor.
• tkvdb.iterators -- pythonic iterators for tkvdb.cursor.
• tkvdb.errors -- all db-related exceptions that code may throw.
• tkvdb.params -- database and transaction params. Wrapper around tkvdb_params.

Database initialization

Database is wrapped into the Tkvdb object from tkvdb or tkvdb.db modules. At this time only path to database file is supported. Parameters (tkvdb.params.Params) optionally may be passed to constructor.

from tkvdb import Tkvdb

db = Tkvdb(path_to_db_file)
# some code
db.close()

Context manager (with statment) that includes auto closing is also available:

with Tkvdb(path_to_db_file) as db:
    # some code
    with db.transaction() as tr:
        # more code

Attributes (readonly):

• path: str -- path to database file.
• is_opened: bool -- shows that database is initialized properly.

Methods (may raise exceptions):

• Tkvdb(path: str, params: tkvdb.params.Params = None) (constructor) -- create database instance.
• close() -- close database.
• transaction(params: tkvdb.params.Params = None) -> tkvdb.transaction.Transaction -- create transaction.

There is also Cython method get_db that returns tkvdb_db * pointer.

Transactions

Transactions are basic way to do any operation with database. Consult with original documentation about transaction term, because it doesn't mean same thing as in other database systems.

Input and ouput uses bytes type for everything. Encode and decode strings if needed.

Parameters (tkvdb.params.Params) optionally may be passed to constructor.

Transaction must be created from database instance (described in previous part):

transaction = db.transaction()
transaction.begin()
transaction.put(b'key', b'value')
transaction.commit() # or transaction.rollback()
print(transaction.getvalue(b'key'))
transaction.free()

Pythonic with statment also available:

with db.transaction() as tr:
    tr.put(b'key', b'value')
    tr.commit()
    print(tr.getvalue(b'key'))

Note that with statement does not do commit, but rollbacks on exception. Do commit or rollback with your own code, or don't do anything (implies rollback-like behavior). Transaction is started (begin) automatically and will be freed (free) at exit from with block though.

Transaction also has Python dict-like interface:

• __getitem__ and __setitem__
• get(key, default=None)
• keys(), values() and items() iterators

with db.transaction() as tr:
    tr[b'key'] = b'value'
    print(tr.get(b'other-key', b'default')) # prints b'default'
    tr.commit()
    print(tr[b'key']) # prints b'value'

    # Iterators
    for key in tr: # or tr.keys()
        print(key)
    for key, value in tr.items():
        print(key, value)
    for value in tr.values():
        print(value)

Multiple keys may be deleted by delete method using optional prefix argument (default False). Passing True allows deleting keys starting from prefix. Dict-like del operator always deletes only exact key.

Attributes (readonly):

• is_initialized: bool -- shows that transaction underlying structures are initialized properly.
• is_started: bool -- shows that begin() method was called.
• is_changed: bool -- shows that transaction had any uncommited changes (i.e. put() was used).
• ram_only: bool -- indicates that transaction is RAM-only.

Transaction methods. Most of them may raise an exception:

• Transaction(ram_only=True, params: tkvdb.params.Params = None) (constructor) -- create transaction instance. Must be called manually only for RAM-only usage, otherwise db.transaction() must be used instead.
• begin() -- starts transaction, calls underlying tkvdb_tr->begin().
• getvalue(key: bytes) -> bytes -- get value by key.
• put(key: bytes) -- insert value into db by key.
• get(key: bytes, default: bytes = None) -> bytes -- dict-like get with default value.
• delete(key: bytes, prefix=False) -- delete value by key. With prefix=True all keys staring with key will be deleted.
• __getitem__, __setitem__, __delitem__ -- dict-like methods.
• __contains__ -- allows usage of in operator.
• free() -- free transaction (called in with statement automatically).
• keys(), values(), items() -- return dict-like iterators.
• cursor(seek_key=None, seek_type=Seek.EQ) -- return transaction cursor (see Cursors), with optional seek.

RAM-only transactions

Transactions also may be used in RAM-only mode. These transactions don't require database file and use less memory. They are cleared on commit() or rollback(). See more about RAM-only transactions in original documentation.

Use tkvdb.Transaction() constructor to create RAM-only transaction without database. This transaction may also be used with with statement, same auto-begin rules apply. Example:

with Transaction() as tr:
    tr[b'key'] = b'value'
    print(tr[b'key'])  # prints b'value'
    tr.commit()  # clears transaction

Iterators

Transaction can be traversed using iterators. It is also the main way for iterating through database contents.

Module tkvdb.iterators provides three dict-like iterators that use tkvdb.cursor.Cursor inside:

• tkvdb.iterators.KeysIterator -- iterating over keys.
• tkvdb.iterators.ValuesIterator -- iterating over values.
• tkvdb.iterators.ItemsIterator -- iterating over key-value pair.

They can be used with transaction:

with db.transaction() as tr:
    for key in tr:  # or tr.keys()
        print(tr[key])
    for value in tr.values():
        print(value)
    for key, value in tr.items():
        print(key, value)

In all loops new instance of Cursor is used.

They also can be used with cursor:

with db.transaction() as tr:
    with tr.cursor() as c:
        for key in c:
            print(c.key(), c.keysize())

Notice that cursor iterators use same underlying Cursor object, so they would iterate from same place where cursor stopped before:

with db.transaction() as tr:
    with tr.cursor() as c:
        c.first()
        # do some iteration with c.next()
        for key in c:
            print(key)  # it wouldn't be first key
            if something:
                break
        for value in c.values():  # starts from last iterated key
            print(value)

Reverse iteration is available through the standard reversed function. Iterators and transaction have required methods for this. For new cursors iteration will start from the end using tkvdb.cursor.Cursor.last.

with db.transaction() as tr:
    for key in reversed(tr):
        print(key)
    # All iterator types allow this
    for key, value in reversed(tr.items()):
        print(key, value)

with db.transaction() as tr:
    with tr.cursor() as c:
        for key in reversed(c):
            print(key)
    with tr.cursor() as c:
        for value in reversed(c.values()):
            print(value)

Cursors created from transacton can used for searching (as shorthand for tkvdb.cursor.Cursor.seek()). More information in next section.

with db.transaction() as tr:
    with tr.cursor(seek_key=b'seek-tr-31', seek_type=Seek.GE) as c:
        print(c.key())

Cursors

Cursors are used to iterate through database contents. They are defined in tkvdb.cursor module, C implementation is wrapped in tkvdb.cursor.Cursor class.

Cursors are attached to transaction and created by Transacion.cursor() method. They also may be created directly.

Although cursors are sole way to iterate and seek in tkvdb, it is better and easier to use python-style iterators for such purposes.

Example usage:

with db.transaction() as tr:
    with tr.cursor() as c:
        c.first()
        while True:
            print(c.key(), c.value())
            try:
                c.next()
            except tkvdb.errors.NotFoundError:
                break

Cursor also may be used without with statement, it would be freed anyway on garbage collection:

with db.transaction() as tr:
    c = tr.cursor():
    c.first()
    # ...

Notice: first and next methods throw tkvdb.errors.EmptyError on empty database, not NotFoundError. Cursors may be iterated by using iterators (see previous section).

Cursor also may be iterated in reverse order using prev() method. Another method called last() moves cursor to last record and often useful for the reverse iteration.

Cursor can be used for search with seek() method. This allows searching k-v pair by prefix using seek criteria. Criterias are defined in tkvdb.cursor.Seek enum:

• Seek.EQ -- search for the exact key match.
• Seek.LE -- search for less (in terms of memcmp()) or equal key.
• Seek.GE -- search for greater (in terms of memcmp()) or equal key.

Seeking is also may be initiated using Transaction.cursor() method (see more in Transactions section).

from tkvdb.cursor import Seek

with db.transaction() as tr:
    with tr.cursor() as c:
        c.seek(b'key', Seek.EQ)
        key = c.key()
        # ...
        c.next()

    with tr.cursor(seek_key=b'key') as c:
        # ...

Attributes (readonly):

• is_initialized: bool -- shows that cursor underlying structures are initialized properly.
• is_started: bool -- shows that first() method was called.

Cursor methods.

• first() -- move cursor to first item in database.
• last() -- move cursor to last item in database.
• next() -- move cursor to next item in database.
• prev() -- move cursor to previous item in database.
• key() -> bytes -- get current key.
• val() -> bytes -- get current value.
• keysize() -> int -- get current key size.
• valsize() -> int -- get current value size.
• free() -- free cursor.
• __iter__() -- returns tkvdb.iterators.KeysIterator.
• seek(key: bytes, seek: tkvdb.cursor.Seek) -- search key by criteria.
• keys(), values(), items() -- return dict-like iterators.

Params

Params are used to specify different options for database and/or transactions. They are defined in tkvdb.params module: the C implementation (tkvdb_params struct) is wrapped by tkvdb.params.Params class, and param values are wrapped by the tkvdb.params.Param enum.

Parameter names transfomed using CamelCase, example:

TKVDB_PARAM_TR_DYNALLOC => TrDynalloc
TKVDB_PARAM_CURSOR_STACK_DYNALLOC => CursorStackDynalloc

Available parameters in tkvdb.params.Param enum:

• TrDynalloc -- TKVDB_PARAM_TR_DYNALLOC
• TrLimit -- TKVDB_PARAM_TR_LIMIT
• Alignval -- TKVDB_PARAM_ALIGNVAL
• Autobegin -- TKVDB_PARAM_AUTOBEGIN
• CursorStackDynalloc -- TKVDB_PARAM_CURSOR_STACK_DYNALLOC
• CursorStackLimit -- TKVDB_PARAM_CURSOR_STACK_LIMIT
• CursorKeyDynalloc -- TKVDB_PARAM_CURSOR_KEY_DYNALLOC
• CursorKeyLimit -- TKVDB_PARAM_CURSOR_KEY_LIMIT
• DbfileOpenFlags -- TKVDB_PARAM_DBFILE_OPEN_FLAGS

Consult with original tkvdb documentation for params meaning and possible values.

Params usage:

from tkvdb.params import Params, Param

# Passing params to database
params = Params({Param.Autobegin: 1})  # set params at init
with Tkvdb(path, params) as db:
    # Params also will be passed to db transactions
    with db.transaction() as tr:
        # ...
    # Transaction may use own params
    params = Params()
    params.set(Param.TrLimit, 100)
    with db.transaction(params) as tr:
        # ...

# Setting params after init
params = Params()
params.set(Param.Autobegin, 1)
tr = Transaction(params=params)  # RAM-only transaction

As in original tkvdb code, params from database are passed to db-bound transaction, if they aren't overrided directly by passing another Params instance to transaction.

Python implementation also stores all set params and stores them in internal values dict. Notice that it tracks only params that were set directly, so default values aren't known to Params wrapper.

Params attributes:

• is_initialized: bool -- shows that params underlying structures are initialized properly.

Params methods:

• Params(params=None) (constructor) -- create params instance. Argument params may be dict with param names and values.
• get_values() -- return all set params.
• get(param: tkvdb.params.Param) -- return single param value.
• set(param: tkvdb.params.Param, value: int) -- set param value.
• free() -- free params object.

There is also Cython method get_params that returns tkvdb_params * pointer.

Errors

Error classes are defined in tkvdb.errors module. Every non-ok return value from the underlying C code is converted to python Exception. Only TKVDB_RES.TKVDB_OK considered as success.

Consult with original documentation for error codes meaning.

One exception from this rule is tkvdb.transaction.Transaction.__getitem__() (dict-like access) that raises KeyError for python compatibility.

Examples:

from tkvdb.errors import EmptyError, NotFoundError, NotStartedError

# ...
tr = db.transaction()
try:
    print(tr.getvalue(b'key'))
except (NotFoundError, EmptyError):
    print('key not found')
except NotStartedError:
    print('transaction not started')

with db.transaction() as tr:
    try:
        print(tr[b'key'])
    except KeyError:
        print('key not found')

Note that tkvdb raises EmptyError (TKVDB_RES.TKVDB_EMPTY return code), not NotFoundError when key is not found in empty database,

Errors:

• Error -- base class for all tkvdb-related errors.
• IoError -- TKVDB_RES.TKVDB_IO_ERROR code.
• LockedError -- TKVDB_RES.TKVDB_LOCKED code.
• EmptyError -- TKVDB_RES.TKVDB_EMPTY code.
• NotFoundError -- TKVDB_RES.TKVDB_NOT_FOUND code.
• EnomemError -- TKVDB_RES.TKVDB_ENOMEM code.
• CorruptedError -- TKVDB_RES.TKVDB_CORRUPTED code.
• NotStartedError -- TKVDB_RES.TKVDB_NOT_STARTED code.
• ModifiedError -- TKVDB_RES.TKVDB_MODIFIED code.

License

Python-tkvdb is licensed under ISC license as original tkvdb project.