pycedar

version python license

Python binding of cedar (implementation of efficiently-updatable double-array trie) using Cython

日本語版の README は README.ja.md にあります。

Official URL of cedar: http://www.tkl.iis.u-tokyo.ac.jp/~ynaga/cedar/

Requirements

Python 3.9 or newer
A 64-bit platform: Windows, or a POSIX-compatible one (Linux, macOS)
A C++ compiler, when building from source (wheels need none)

The extension is tested with CPython 3.9 through 3.14 on Linux, macOS and Windows, including the free-threaded build of 3.14 (3.14t). Wheels cover Linux (x86_64, aarch64), macOS (x86_64, arm64) and Windows (AMD64, ARM64). The module is declared free-threading compatible, so it imports and works on a free-threaded interpreter without a ModuleNotFoundError; concurrency across distinct trie objects is supported (see Limitations).

Why a double-array trie?

A trie encodes the keys into its arrays, so it does not keep a separate string object per key, and it answers prefix queries that a hash map cannot.

200000 random lowercase keys of 4 to 16 characters, mapped to integers, measured with benchmarks/bench.py on CPython 3.13 (Linux, x86_64):

	builtin `dict`	pycedar
hash table and values	13.43 MB	—
key strings	9.73 MB	—
node array	—	2.35 MB
tail array	—	2.06 MB
total	23.16 MB	4.40 MB
per key	121 bytes	23 bytes

The trade is real and worth stating plainly: a point lookup costs roughly twice what a dict lookup does (30 ns against 16 ns on the same data; the numbers vary by machine). Reach for pycedar when the key set is large enough that memory matters, or when you need prefix search; reach for a dict when you only need point lookups.

Run python benchmarks/bench.py --help to reproduce these numbers on your own data.

Installation

install from PyPI release

$ pip install --user pycedar

install from GitHub master

$ pip install --user https://github.com/akivajp/pycedar/archive/master.zip

Type checking

Type stubs ship as the pycedar-stubs package (PEP 561). Type checkers and IDEs pick them up automatically once pycedar is installed — no extra step. (型スタブは pycedar-stubs パッケージとして同梱されるため、インストール後は型検査器や IDE が自動で API を認識する)

Usage

using python-like dict class based on double array trie

>>> import pycedar

>>> d = pycedar.dict()
>>> len(d)
0
>>> bool(d)
False
>>> list(d)
[]

>>> d['nineteen'] = 19
>>> d.set('twenty', 20)
20
>>> d['twenty one'] = 21
>>> d['twenty two'] = 22
>>> d['twenty three'] = 23
>>> d['twenty four'] = 24

>>> len(d)
6
>>> bool(d)
True
>>> list(d)
['nineteen', 'twenty', 'twenty four', 'twenty one', 'twenty three', 'twenty two']
>>> list(d.keys())
['nineteen', 'twenty', 'twenty four', 'twenty one', 'twenty three', 'twenty two']
>>> list(d.values())
[19, 20, 24, 21, 23, 22]
>>> list(d.items())
[('nineteen', 19), ('twenty', 20), ('twenty four', 24), ('twenty one', 21), ('twenty three', 23), ('twenty two', 22)]
>>> d['twenty four']
24
>>> 'twenty four' in d
True
>>> del d['twenty four']
>>> 'twenty four' in d
False
>>> d['twenty four']
Traceback (most recent call last):
    ...
KeyError: 'twenty four'
>>> d.get('twenty three')
23
>>> d.get('twenty four')       # the default is base_trie.NO_VALUE
-1
>>> d.get('twenty four', None) is None
True

Prefix queries return generators:

>>> list(d.find(''))
[('nineteen', 19), ('twenty', 20), ('twenty one', 21), ('twenty three', 23), ('twenty two', 22)]
>>> list(d.find('tw'))
[('twenty', 20), ('twenty one', 21), ('twenty three', 23), ('twenty two', 22)]
>>> list(d.find('twenty t'))
[('twenty three', 23), ('twenty two', 22)]
>>> list(d.find_keys('twenty'))
['twenty', 'twenty one', 'twenty three', 'twenty two']
>>> list(d.find_values('twenty'))
[20, 21, 23, 22]

Nodes let you keep a position in the trie and search relative to it:

>>> n = d.get_node('twenty')
>>> n.key()
'twenty'
>>> n.value()
20
>>> [child.key() for child in n.find_nodes(' t')]
[' three', ' two']

>>> d.get_node('twenty ') is None    # a path without a value
True

Saving and loading:

>>> d.save('test.dat')
0
>>> d2 = pycedar.dict()
>>> d2.setdefault('eighteen', 18)
18
>>> list(d2.items())
[('eighteen', 18)]
>>> d2.load('test.dat')              # replaces the whole trie image
0
>>> list(d2.items())
[('nineteen', 19), ('twenty', 20), ('twenty one', 21), ('twenty three', 23), ('twenty two', 22)]
>>> d2.setdefault('eighteen', 18)
18
>>> list(d2.items())
[('eighteen', 18), ('nineteen', 19), ('twenty', 20), ('twenty one', 21), ('twenty three', 23), ('twenty two', 22)]

The image can also be kept in memory as bytes, and the dict supports pickling and copying:

>>> import pickle, copy

>>> image = d.dumps()
>>> restored = pickle.loads(pickle.dumps(d))
>>> restored['twenty']
20
>>> clone = copy.deepcopy(d)             # copy.copy works the same way
>>> clone['twenty'] = 999
>>> d['twenty']
20

>>> d.pop('twenty two')                  # like dict.pop
22
>>> d.pop('twenty two', 'gone')
'gone'
>>> d.popitem()                          # removes the first key in sorted order
('nineteen', 19)

popitem() returns the lexicographically smallest remaining key, because a trie keeps no insertion order — this is the one place pycedar.dict intentionally differs from dict.

using bytes keys

pycedar.dict is parameterised by its key type, which is enforced strictly:

>>> b = pycedar.dict(bytes)
>>> b[b'cedar'] = 1
>>> b[b'cedarpp'] = 2
>>> list(b.items())
[(b'cedar', 1), (b'cedarpp', 2)]
>>> b['str key'] = 3
Traceback (most recent call last):
    ...
TypeError: Argument 'key' has incorrect type (expected bytes, got str)

str keys are encoded as UTF-8 before they reach cedar, so lengths reported by the low level API are byte lengths, not character counts.

using more primitive data structures

pycedar.dict is a thin convenience layer over the trie classes. You can use them directly when you want cedar's raw semantics.

>>> t = pycedar.str_trie()
>>> t.set('apple', 1)
1
>>> t.set('applet', 2)
2
>>> t.set('apply', 3)
3

>>> t.exact_match_search('apple')      # (value, length, node id)
(1, 5, 259)
>>> t.exact_match_search('app')[0]     # a prefix carries no value
-1

>>> t.common_prefix_search('applet')   # every key that prefixes the query
[('apple', 1, 259), ('applet', 2, 368)]
>>> [key for key, value, node_id in t.common_prefix_predict('app')]
['le', 'let', 'ly']

>>> t.erase('apply')
0
>>> t.erase('apply')                   # already gone
-1
>>> t.num_keys()
2

Note the difference between the two prefix queries:

common_prefix_search(key) returns the complete keys that are prefixes of key.
common_prefix_predict(key) returns the remaining suffixes of every key that starts with key.

Both also have lazy variants that yield one triple at a time instead of building a list, which matters when a query can match a large part of the trie or when the first hits are all you need:

>>> [key for key, value, node_id in t.icommon_prefix_predict('app')]   # 'apply' was erased above
['le', 'let']
>>> list(t.icommon_prefix_search('applet'))
[('apple', 1, 259), ('applet', 2, 368)]

The lazy variants return the same triples in the same order as the list versions; results are not guaranteed if the trie is modified while iterating. Whether laziness pays depends on how early you stop. Measured with benchmarks/bench.py on the 20000-key set from the table above, taking only the first hit of a 2-character predict costs roughly a tenth of the full list (332 ns against 2892 ns per prefix), while draining the whole generator costs about 10% more than the list version; the search variants, whose results are capped at one per key byte, are cheap either way. (遅延版は list 版と同じ組を同じ順で返す。反復中のトライ変更時の結果は保証されない。早期に打ち切るほど遅延版が有利になる。上の表と同じ 20000 キーで benchmarks/bench.py により計測した結果、2 文字の predict で先頭 1 件のみの取得は全件 list 化の約 1/10 (プレフィックスあたり 332 ns に対し 2892 ns)、ジェネレータを最後まで消費すると list 版より約 1 割遅い。search の結果はキー1バイトあたり最大1件のためどちらでも安価)

Enumerating a trie (or a subtree) uses begin / next:

>>> result, from_id, pos = t.traverse('app')   # locate the subtree
>>> value, node_id, length = t.begin(from_id, pos)
>>> while value != pycedar.base_trie.NO_PATH:
...     print(t.suffix(node_id, length), value)
...     value, node_id, length = t.next(node_id, length, from_id)
apple 1
applet 2

API reference

`pycedar.base_trie`

Base class for every trie. Not meant to be instantiated directly.

Member	Description
`NO_VALUE`	`-1`. Returned when a node exists but carries no value.
`NO_PATH`	`-2`. Returned when the path does not exist, and used as the traversal terminator.
`root`	The `node` object for the trie root.
`clear(reuse=True)`	Drop all keys.
`capacity()`, `size()`, `length()`, `total_size()`, `unit_size()`, `nonzero_size()`, `nonzero_length()`, `num_keys()`	cedar's internal statistics. `num_keys()` is the number of registered keys.
`begin(from_id=0, length=0)`	Start an enumeration. Returns `(value, node_id, length)`.
`next(node_id, length, root=0)`	Advance an enumeration. Returns `(value, node_id, length)`.
`open(filepath, mode='rb', offset=0, size=0)`	Load a trie image. Returns `0` on success, `-1` on failure.
`save(filepath, mode='wb', shrink=True)`	Write a trie image. Returns `0` on success, `-1` on failure.
`dumps(shrink=True)`	The trie image as `bytes`. The layout is byte-for-byte what `save()` writes, so the two are interchangeable.
`loads(data)`	Replace the trie with an image from any bytes-like object (`bytes`, `bytearray`, `memoryview`, ...). Returns `0` / `-1`. Unlike a failed `load()`, malformed input is rejected before the current contents are touched, so the trie keeps them.
`icommon_prefix_search(key, from_id=0, max_size=-1)`	Lazy generator variant of `common_prefix_search()`; yields the same `(key, value, node_id)` triples one at a time. Inherited by the specializations.
`icommon_prefix_predict(key, from_id=0, max_size=-1)`	Lazy generator variant of `common_prefix_predict()`; yields the same `(suffix, value, node_id)` triples one at a time. Inherited by the specializations.

`pycedar.str_trie` / `pycedar.bytes_trie` / `pycedar.unicode_trie`

Specialisations of base_trie for str keys, bytes keys, and unicode keys respectively. On Python 3, unicode is str, so unicode_trie behaves identically to str_trie and is kept only for backward compatibility.

In addition to the base_trie members:

Method	Description
`set(key, value)`	Register `key` with `value`. Returns the stored value. Raises `KeyError` for an empty key.
`update(key, delta=0)`	Register `key` if needed and add `delta` to its value. Returns the new value.
`erase(key, from_id=0)`	Remove `key`. Returns `0` on success, `-1` if it was not registered.
`exact_match_search(key, from_id=0)`	Returns `(value, length, node_id)`. `value` is `NO_VALUE` / `NO_PATH` when not found.
`common_prefix_search(key, from_id=0, max_size=-1)`	Returns a list of `(key, value, node_id)` for every key that prefixes `key`.
`common_prefix_predict(key, from_id=0, max_size=-1)`	Returns a list of `(suffix, value, node_id)` for every key starting with `key`.
`traverse(key, from_id=0, pos=0)`	Follow `key` from `from_id`. Returns `(value, node_id, pos)`.
`suffix(node_id, length=0)`	Reconstruct the key ending at `node_id`.

max_size caps the number of returned results; -1 means "no limit". from_id scopes the query to a subtree.

`pycedar.node`

A cursor into a trie. Obtained from base_trie.root, dict.root, dict.get_node(), dict.nodes() or node.find_nodes().

Member	Description
`id`, `length`, `root`	Read-only position information.
`key()`	The key this node represents, relative to `root`.
`value()`	The value stored at this node.
`track()`	Returns `(id, length, root)`.
`traverse(key)`	Generator yielding `(value, node_id, length)` for the subtree under `key`.
`find_nodes(key)`	Generator yielding `node` objects for the subtree under `key`.
`get_node(key)`	The `node` for `key`, or `None` if it is absent or carries no value.

`pycedar.dict`

A dict-like façade over a trie. pycedar.dict(key_type) accepts str (the default) or bytes.

Member	Description
`trie`, `root`, `type`	The underlying trie, its root `node`, and the key type.
`d[key]`	The value, or `KeyError`.
`d[key] = value`	Register a key. `KeyError` for an empty key.
`del d[key]`	Remove a key, or `KeyError`.
`key in d`, `len(d)`, `iter(d)`	Membership, number of keys, iteration over keys.
`get(key, default=NO_VALUE)`	The value, or `default`.
`set(key, value)`	Register a key. Returns the stored value.
`setdefault(key, value=0)`	Register only if absent. Returns the effective value.
`pop(key, default=…)`	Remove `key` and return its value; raises `KeyError` without a default when absent.
`popitem()`	Remove and return the first `(key, value)` in sorted-key order. `KeyError` when empty. Unlike `dict.popitem()`, which pops the most recent insertion — a trie keeps no insertion order.
`update(key, delta=0)`	Add `delta` to a key's value, registering it if needed.
`clear()`	Drop all keys.
`keys()`, `values()`, `items()`, `nodes()`	Generators over the whole trie.
`find(prefix)`, `find_keys(prefix)`, `find_values(prefix)`	Generators scoped to `prefix`.
`get_node(key)`	The `node` for `key`, or `None`.
`save(filepath, mode='wb', shrink=True)`	Write a trie image. Returns `0` / `-1`.
`load(filepath, mode='rb')`	Replace the trie with a stored image. Returns `0` / `-1`.
`dumps(shrink=True)`	The trie image as `bytes`; interchangeable with `save()` files.
`loads(data)`	Replace the trie with an image from `dumps()`/`save()`, given as any bytes-like object (`bytes`, `bytearray`, `memoryview`, ...). Returns `0` / `-1`.

pycedar.dict (and the trie classes) support pickle, copy.copy() and copy.deepcopy() through the same serialized image.

pycedar.__version__ exposes the installed package version.

Limitations

Concurrent access is safe only across distinct tries

The module is declared free-threading compatible (PEP 703) and works on free-threaded CPython 3.14. That covers using separate trie objects from several threads at once, which the test suite exercises.

It does not make an individual trie thread-safe: concurrent operations on the same trie race, exactly as they would under the GIL, and can corrupt it. Guard a shared trie with a lock as you would any other mutable object.

Values are C `int` sized, and two of them are reserved

Values are stored as C int, so they must fit in -2**31 .. 2**31-1; anything larger raises OverflowError.

-1 and -2 are cedar's sentinels — base_trie.NO_VALUE and base_trie.NO_PATH — and every writer rejects them:

>>> limited = pycedar.dict()
>>> limited['key'] = -1
Traceback (most recent call last):
    ...
ValueError: -1 is reserved and cannot be stored: ...

update() checks the resulting value rather than the delta, because a perfectly ordinary delta can still land on a sentinel. When it does, the delta is rolled back and the stored value is left untouched:

>>> limited['counter'] = 1
>>> limited.update('counter', -3)     # 1 + (-3) == -2
Traceback (most recent call last):
    ...
ValueError: -2 is reserved and cannot be stored: ...
>>> limited['counter']
1

Every other value round trips, negative ones included.

Before 0.3.0 these two were accepted and silently corrupted the trie: -1 made a key invisible to in, get() and d[key] while leaving it visible to iteration, and -2 ended every traversal early, hiding each key that came after it. If you are upgrading and were storing either, the values were not being read back correctly in the first place.

Keys cannot contain a NUL byte

cedar keeps short key suffixes in a NUL terminated array, so a key carrying a NUL of its own breaks that invariant. Such keys are rejected:

>>> keyed = pycedar.dict()
>>> keyed['a\x00b'] = 1
Traceback (most recent call last):
    ...
ValueError: key contains a NUL byte, ...

Before 0.4.0 the write was accepted, and it did more than read back wrong: inserting such a key and then inserting one that shared its prefix corrupted memory and crashed the interpreter.

The serialization format is platform-dependent and unauthenticated

The native cedar .dat format depends on the pointer size and byte order of the machine that wrote it, and it carries no integrity checks. Only load files produced by pycedar on a compatible platform and obtained from a trusted source.

The same applies to dumps() / loads(), which use the identical layout, and to pickle — pickling serializes this image, so pickles are only portable between machines of the same platform and integrity is not verified. (シリアライズはプラットフォーム依存かつ改竄検査なし。dumps()/loads() や pickle も同一レイアウトを使用するため同様)

I/O failures are reported through return codes

save() / load() / open() return 0 on success and -1 on failure instead of raising; check the return value.

Running out of memory is the exception to that rule: save() and load() raise MemoryError rather than returning -1. A failed load() leaves the trie empty, because the previous contents are released before the new ones are allocated. The instance stays valid and can be reused. See pycedar/core/cedar/README.md.

Development

$ python -m pip install --upgrade pip
$ python -m pip install . pytest
$ pytest

To rebuild in place while iterating on pycedar.pyx:

$ python -m pip install "Cython>=3.1,<4" setuptools
$ python setup.py build_ext --inplace

./clean.sh removes build artifacts.

Benchmarks

benchmarks/bench.py times the operations pycedar is used for. Run it against two builds to check whether a change actually paid off:

$ python benchmarks/bench.py --label before
$ python benchmarks/bench.py --label after

It uses rich for the table when that is installed, and plain text otherwise. --help lists the knobs.

Releasing

The version lives in a single place, pycedar/VERSION.

Update pycedar/VERSION and CHANGELOG.md.
Commit and push to master.
Push a matching tag, e.g. git tag v0.2.0 && git push origin v0.2.0.

The release workflow verifies that the tag matches pycedar/VERSION, builds the sdist and the Linux/macOS wheels with cibuildwheel, and publishes them to PyPI via Trusted Publishing.

License

pycedar is distributed under the same terms as cedar itself: GPLv2, LGPLv2.1 and BSD-2-Clause. See the license files bundled under pycedar/core/cedar/.

Credits

cedar is written by Naoki Yoshinaga. The copy vendored under pycedar/core/cedar/ carries local modifications. pycedar/core/cedar/README.md records what was changed and why, why the 2022 upstream tarball was deliberately not re-vendored, and what has to be re-applied if anyone syncs with a newer release.

See CHANGELOG.md for the list of contributors to each release.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pycedar-0.6.1.tar.gz (375.1 kB view details)

Uploaded Sep 18, 2026 Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

pycedar-0.6.1-cp314-cp314t-win_arm64.whl (589.3 kB view details)

Uploaded Sep 18, 2026 CPython 3.14tWindows ARM64

pycedar-0.6.1-cp314-cp314t-win_amd64.whl (441.2 kB view details)

Uploaded Sep 18, 2026 CPython 3.14tWindows x86-64

pycedar-0.6.1-cp314-cp314t-musllinux_1_2_x86_64.whl (2.5 MB view details)

Uploaded Sep 18, 2026 CPython 3.14tmusllinux: musl 1.2+ x86-64

pycedar-0.6.1-cp314-cp314t-musllinux_1_2_aarch64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.14tmusllinux: musl 1.2+ ARM64

pycedar-0.6.1-cp314-cp314t-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (1.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.14tmanylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

pycedar-0.6.1-cp314-cp314t-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (1.5 MB view details)

Uploaded Sep 18, 2026 CPython 3.14tmanylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

pycedar-0.6.1-cp314-cp314t-macosx_11_0_arm64.whl (282.5 kB view details)

Uploaded Sep 18, 2026 CPython 3.14tmacOS 11.0+ ARM64

pycedar-0.6.1-cp314-cp314t-macosx_10_15_x86_64.whl (288.8 kB view details)

Uploaded Sep 18, 2026 CPython 3.14tmacOS 10.15+ x86-64

pycedar-0.6.1-cp314-cp314-win_arm64.whl (580.5 kB view details)

Uploaded Sep 18, 2026 CPython 3.14Windows ARM64

pycedar-0.6.1-cp314-cp314-win_amd64.whl (430.7 kB view details)

Uploaded Sep 18, 2026 CPython 3.14Windows x86-64

pycedar-0.6.1-cp314-cp314-musllinux_1_2_x86_64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.14musllinux: musl 1.2+ x86-64

pycedar-0.6.1-cp314-cp314-musllinux_1_2_aarch64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.14musllinux: musl 1.2+ ARM64

pycedar-0.6.1-cp314-cp314-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (1.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.14manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

pycedar-0.6.1-cp314-cp314-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (1.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.14manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

pycedar-0.6.1-cp314-cp314-macosx_11_0_arm64.whl (274.5 kB view details)

Uploaded Sep 18, 2026 CPython 3.14macOS 11.0+ ARM64

pycedar-0.6.1-cp314-cp314-macosx_10_15_x86_64.whl (282.7 kB view details)

Uploaded Sep 18, 2026 CPython 3.14macOS 10.15+ x86-64

pycedar-0.6.1-cp313-cp313-win_arm64.whl (562.4 kB view details)

Uploaded Sep 18, 2026 CPython 3.13Windows ARM64

pycedar-0.6.1-cp313-cp313-win_amd64.whl (418.8 kB view details)

Uploaded Sep 18, 2026 CPython 3.13Windows x86-64

pycedar-0.6.1-cp313-cp313-musllinux_1_2_x86_64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.13musllinux: musl 1.2+ x86-64

pycedar-0.6.1-cp313-cp313-musllinux_1_2_aarch64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.13musllinux: musl 1.2+ ARM64

pycedar-0.6.1-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (1.5 MB view details)

Uploaded Sep 18, 2026 CPython 3.13manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

pycedar-0.6.1-cp313-cp313-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (1.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.13manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

pycedar-0.6.1-cp313-cp313-macosx_11_0_arm64.whl (272.5 kB view details)

Uploaded Sep 18, 2026 CPython 3.13macOS 11.0+ ARM64

pycedar-0.6.1-cp313-cp313-macosx_10_13_x86_64.whl (281.8 kB view details)

Uploaded Sep 18, 2026 CPython 3.13macOS 10.13+ x86-64

pycedar-0.6.1-cp312-cp312-win_arm64.whl (562.5 kB view details)

Uploaded Sep 18, 2026 CPython 3.12Windows ARM64

pycedar-0.6.1-cp312-cp312-win_amd64.whl (419.0 kB view details)

Uploaded Sep 18, 2026 CPython 3.12Windows x86-64

pycedar-0.6.1-cp312-cp312-musllinux_1_2_x86_64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.12musllinux: musl 1.2+ x86-64

pycedar-0.6.1-cp312-cp312-musllinux_1_2_aarch64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.12musllinux: musl 1.2+ ARM64

pycedar-0.6.1-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (1.5 MB view details)

Uploaded Sep 18, 2026 CPython 3.12manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

pycedar-0.6.1-cp312-cp312-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (1.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.12manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

pycedar-0.6.1-cp312-cp312-macosx_11_0_arm64.whl (270.3 kB view details)

Uploaded Sep 18, 2026 CPython 3.12macOS 11.0+ ARM64

pycedar-0.6.1-cp312-cp312-macosx_10_13_x86_64.whl (276.9 kB view details)

Uploaded Sep 18, 2026 CPython 3.12macOS 10.13+ x86-64

pycedar-0.6.1-cp311-cp311-win_arm64.whl (566.5 kB view details)

Uploaded Sep 18, 2026 CPython 3.11Windows ARM64

pycedar-0.6.1-cp311-cp311-win_amd64.whl (421.2 kB view details)

Uploaded Sep 18, 2026 CPython 3.11Windows x86-64

pycedar-0.6.1-cp311-cp311-musllinux_1_2_x86_64.whl (2.5 MB view details)

Uploaded Sep 18, 2026 CPython 3.11musllinux: musl 1.2+ x86-64

pycedar-0.6.1-cp311-cp311-musllinux_1_2_aarch64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.11musllinux: musl 1.2+ ARM64

pycedar-0.6.1-cp311-cp311-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (1.5 MB view details)

Uploaded Sep 18, 2026 CPython 3.11manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

pycedar-0.6.1-cp311-cp311-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (1.5 MB view details)

Uploaded Sep 18, 2026 CPython 3.11manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

pycedar-0.6.1-cp311-cp311-macosx_11_0_arm64.whl (279.1 kB view details)

Uploaded Sep 18, 2026 CPython 3.11macOS 11.0+ ARM64

pycedar-0.6.1-cp311-cp311-macosx_10_9_x86_64.whl (288.3 kB view details)

Uploaded Sep 18, 2026 CPython 3.11macOS 10.9+ x86-64

pycedar-0.6.1-cp310-cp310-win_amd64.whl (420.8 kB view details)

Uploaded Sep 18, 2026 CPython 3.10Windows x86-64

pycedar-0.6.1-cp310-cp310-musllinux_1_2_x86_64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.10musllinux: musl 1.2+ x86-64

pycedar-0.6.1-cp310-cp310-musllinux_1_2_aarch64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.10musllinux: musl 1.2+ ARM64

pycedar-0.6.1-cp310-cp310-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (1.5 MB view details)

Uploaded Sep 18, 2026 CPython 3.10manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

pycedar-0.6.1-cp310-cp310-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (1.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.10manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

pycedar-0.6.1-cp310-cp310-macosx_11_0_arm64.whl (279.2 kB view details)

Uploaded Sep 18, 2026 CPython 3.10macOS 11.0+ ARM64

pycedar-0.6.1-cp310-cp310-macosx_10_9_x86_64.whl (288.1 kB view details)

Uploaded Sep 18, 2026 CPython 3.10macOS 10.9+ x86-64

pycedar-0.6.1-cp39-cp39-win_amd64.whl (422.0 kB view details)

Uploaded Sep 18, 2026 CPython 3.9Windows x86-64

pycedar-0.6.1-cp39-cp39-musllinux_1_2_x86_64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.9musllinux: musl 1.2+ x86-64

pycedar-0.6.1-cp39-cp39-musllinux_1_2_aarch64.whl (2.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.9musllinux: musl 1.2+ ARM64

pycedar-0.6.1-cp39-cp39-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (1.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.9manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

pycedar-0.6.1-cp39-cp39-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (1.4 MB view details)

Uploaded Sep 18, 2026 CPython 3.9manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

pycedar-0.6.1-cp39-cp39-macosx_11_0_arm64.whl (280.0 kB view details)

Uploaded Sep 18, 2026 CPython 3.9macOS 11.0+ ARM64

pycedar-0.6.1-cp39-cp39-macosx_10_9_x86_64.whl (289.0 kB view details)

Uploaded Sep 18, 2026 CPython 3.9macOS 10.9+ x86-64

File details

Details for the file pycedar-0.6.1.tar.gz.

File metadata

Download URL: pycedar-0.6.1.tar.gz
Upload date: Sep 18, 2026
Size: 375.1 kB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for pycedar-0.6.1.tar.gz
Algorithm	Hash digest
SHA256	`e67933eb09c2f416cd5ba1b36d858c021ffd7bd36a68f8b53e0068cf74d43993`
MD5	`3374e493a8b0386b6fa9934f9a70538a`
BLAKE2b-256	`49f633edd29ad3b3f69676973010f40b9b83a96a0a27b031997210b567754562`

Algorithm	Hash digest
SHA256	`03329b36c493abd1d46de3a15ad11221b606c7aba909ef78799418f403ddf3b3`
MD5	`0c09e62bbbf6cac60abd2d23fd77583d`
BLAKE2b-256	`d8f56dc0ca7502998122526c49c54eca00d35616a915e8399709e6d2c759bddd`

Algorithm	Hash digest
SHA256	`58c6957f1990a5bfa67af277582791bbc4fd086cc5689d4e10f830e7a05fe4f7`
MD5	`9ede6c7d9cbd5f98beb0d2b866c8332d`
BLAKE2b-256	`0ccf106b7a92badbca917f2da2325ff30875b359c62c1f281e48776e85397c4e`

Algorithm	Hash digest
SHA256	`f58626104ca00c40bda0387c8fb3fd564ca7f662c17588449820afcd868baeae`
MD5	`11fdfb1b91047c2811f0578c526b51f1`
BLAKE2b-256	`79cda3ce9bc0820032a25387cf3fb939fd3b62b5ab2255700f15f0355bd2347f`

Algorithm	Hash digest
SHA256	`85b245b04db0e6b82f20a8416c7d473c01c749a7431349074b40bcb58a826d0f`
MD5	`ad8013e893b335eef6bdbf1c0fad3af2`
BLAKE2b-256	`4c3e78363664330120c629f1fb46e254e722e03e8cb0f798a7e9e9d4b1460232`

Algorithm	Hash digest
SHA256	`6efc98ae68c86d38156bd8bedcaca0643748601e53fff19405ba0753c106325b`
MD5	`fc8c96a302dacc88d591f3211c28b45d`
BLAKE2b-256	`8f204fe51b91b90e27fc7f99c711494e87fc289ac75713a9adc7944135b9a29a`

Algorithm	Hash digest
SHA256	`caef68790235bb21018db3189f03308c68003122885693d1008803de8e7e983c`
MD5	`f1b4e8f103e08928a9bf06ccd7fdfc6f`
BLAKE2b-256	`700d75b11825db3109540fd29e2b78768587f944b975985342df2deef0078c4f`

Algorithm	Hash digest
SHA256	`329538a26259814ffa9a1bbd464f489e3c62ae820479041243ab1d64e74d36c1`
MD5	`570ad02420de8042a81544a8187c8899`
BLAKE2b-256	`42e6178afb6279c4b18e8eb94c850098213ac69d183a3a8552166f8d8f7d2563`

Algorithm	Hash digest
SHA256	`b723d3cb608af2b7669b7cd705e70c4aef7088b43b73cbe9e03a35764f7dc4bc`
MD5	`f3f79a98c14f5739d8a090d1bc1f610f`
BLAKE2b-256	`2821e5316d6f5bd15324e8f7b6692e4b7bd0745b1e6bf9eb261652cd8618751e`

Algorithm	Hash digest
SHA256	`a2e866a373fa95b038cd3ced06aa16b50b119f6b3ed1e0fad3af670d78419963`
MD5	`db20f6eb70e4036ecce07dd2bf48fc70`
BLAKE2b-256	`773624ded9a70d9d499d41d49376b2459975e13f9d44b6e0ee335c384cd8f39e`

Algorithm	Hash digest
SHA256	`2bb5526127b3d605898604aa8cbbfaef082d980bcbbdaad3ccbc600267c567f4`
MD5	`20c05ed80a3ea4b1104f21e94f0e3f91`
BLAKE2b-256	`50baad7f6f54e615076f1717551eeb582d267f9b7dd722dbcf5d1b3ebf2fede9`

Algorithm	Hash digest
SHA256	`764c931a21017b6c153e4fec11a4fb940f5f625918955b958a26ca7d7dc1f897`
MD5	`fb30777ea1536fee6a4077785685d698`
BLAKE2b-256	`c19bf0ed255c513f9b3d8121bce4d2f01a3917e2b02387a34469f69e0af9650e`

Algorithm	Hash digest
SHA256	`47b67dca6e8440f3d680a9998fdaf98052765b2cce5fe62d70dacf2d085c38b8`
MD5	`324dcd06f7ca84cfe3a86246784d27c9`
BLAKE2b-256	`345691956f27c6e3481f5cb9cc0900a84f9c67f61c62388fd0edbf225b45814e`

Algorithm	Hash digest
SHA256	`012da73006ab36b6691adf459f5c017ded54a2ce3445a875005ebf32a730c9d6`
MD5	`ebb88e26c2f6d33fbecdbe672fea5ed9`
BLAKE2b-256	`5fb269a3ca80cf278eed3ac194c62260597dbe8cd0ce900b904e60a932bd2589`

Algorithm	Hash digest
SHA256	`e8a3f4de5d73017cffb66a7c96af9922f30d049afa8a7085fc38defce7ac430a`
MD5	`6b06a5e646a91d3dceda628e2f62287a`
BLAKE2b-256	`8fb9d1c6394f1fb411fd6d23572882d3bfe3955e22f87d9587e6c23d136cd016`

Algorithm	Hash digest
SHA256	`6f22b94eeb1be6915d074a27400e6fd6041df3c587c23cb7a62c5cec39ce0887`
MD5	`89de1fd5b721d106a88f8ae29a72733a`
BLAKE2b-256	`134892f2c23b7393774b978b64aa4fc36305b08e80b9987d5b8c551907500aee`

Algorithm	Hash digest
SHA256	`0cede32b0e3a3a4339a5a3865470b972c12db2f92345caded8deff0f369a611b`
MD5	`ee2d25dbe4db5636a298761e5335d721`
BLAKE2b-256	`a067d5813339b6accefde93b4be1fe1ef6ef950feafde21530b249c9b0cbbbe8`

Algorithm	Hash digest
SHA256	`8bc4036ef059e84478608c744340b27b677f50fb989feace3869eae0a2ca8dd6`
MD5	`ef6bb5a196f9500b321996dc7e151231`
BLAKE2b-256	`53cfa8c4caaf6e8231f28ce29cf5c5a1a5da91300aa5df8c8482b145f7457b5d`

Algorithm	Hash digest
SHA256	`1af4b4f6ac4d1b19038cc0c08f0c087f2096995684247bda0f596d8717a38f5a`
MD5	`e1161a15acabf615e6d1022e152fe101`
BLAKE2b-256	`26fe53e43fa66f24a9a830daa9606824135496e624d1e962ce22c4628e04e42f`

Algorithm	Hash digest
SHA256	`b697533eaeae2e0fb60e7bb753d620bf03767493c64ca01432756c9aa29a2c0f`
MD5	`98521600cf8f06acc5f60873a69d42d3`
BLAKE2b-256	`6566a07e4fb6c17d460b19ecc964feab0d1978cb99cde6963496a026907b75cf`

Algorithm	Hash digest
SHA256	`4f3930c7ca1d0c0f3f97d1fe686bc4a55e9de834d66879128d67598426f2511d`
MD5	`b8103cda6a7d47055d0801fc82562c07`
BLAKE2b-256	`52601738dc1a60228ab65f918216e2bde360c42b869bd40194897f0665e60076`

pycedar 0.6.1

pycedar

Requirements

Why a double-array trie?

Installation

install from PyPI release

install from GitHub master

Type checking

Usage

using python-like dict class based on double array trie

using bytes keys

using more primitive data structures

API reference

pycedar.base_trie

pycedar.str_trie / pycedar.bytes_trie / pycedar.unicode_trie

pycedar.node

pycedar.dict

Limitations

Concurrent access is safe only across distinct tries

Values are C int sized, and two of them are reserved

Keys cannot contain a NUL byte

The serialization format is platform-dependent and unauthenticated

I/O failures are reported through return codes

Development

Benchmarks

Releasing

License

Credits

Project links

Key dates

1 maintainer

Credits

GitHub Statistics

License

Requires

Classifiers

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

`pycedar.base_trie`

`pycedar.str_trie` / `pycedar.bytes_trie` / `pycedar.unicode_trie`

`pycedar.node`

`pycedar.dict`

Values are C `int` sized, and two of them are reserved

Algorithm	Hash digest
SHA256	`16ec76036fcd15332e2d9cbd0018fb7d3fa23b703bbaaf2331c3bc4657e590ed`
MD5	`7dcbfe378c4713f182e4fd5192f08348`
BLAKE2b-256	`dc081eb98cb6865e8092dda82f5ad43f68ada7a23dd32f4825657fd27f798d86`

Algorithm	Hash digest
SHA256	`1906ccf65120c9b3e08a4a7b82efdb0866ddfa4c56a01210d2e45fa3bccb17e8`
MD5	`fde0548a4323fc9a620c17972f4705e4`
BLAKE2b-256	`781f5b6cdfc2cb7bfd11d3a81243551114555b4a4bc51f174fef97c2e46274b8`

Algorithm	Hash digest
SHA256	`e5cd9411b5739417496ffb8b8d185befbc11b81c50cc08d93107a2212606009f`
MD5	`99716476cf6207ecff270d7cf0fc428b`
BLAKE2b-256	`a2e4da8db7ccd2ca5e48a379715fc4215fde7a3f8279090caaabd390bed2ce10`

Algorithm	Hash digest
SHA256	`f87bef54158215f0e55ef06d1e434a4a91390c0a0971b3b9f9f265e90b0c321e`
MD5	`8570548105c41901f7d0766599086fb3`
BLAKE2b-256	`af4b8546a185384a7298bdf1cc6293c45c1d159bb9cb6adce11eed865be4d778`

Algorithm	Hash digest
SHA256	`9ca31b481f1fcd09d6df84c9b5efe9c9db1d845183c29b8094e6d47b369b5416`
MD5	`eb18812d8d49808d0d97de565efbafdf`
BLAKE2b-256	`db2028c8afae60ef66a022f3be590855365219d94fa7c5e9c088ccefdc167aae`

Algorithm	Hash digest
SHA256	`45e78c41785e426d6f1b80cbe4ddbd0c4f60c4ab2147ff1349fdebdd91cdd94e`
MD5	`0111a0f06a4ed6b02d989f155e44e835`
BLAKE2b-256	`605f09912c2bf3df9a98ef4b33c5019b8d7a8c3177d223f7800746743402f8ae`

Algorithm	Hash digest
SHA256	`c6cf06babc1bd8bc96a86030ae83c3d7b0830aa7af5eb845b42572398423abee`
MD5	`481057bb918036d482384f51d59a429d`
BLAKE2b-256	`7698f2174fa83f6199466ec25b4916209d66e4f15b61812b55ea7652c8b6d062`

Algorithm	Hash digest
SHA256	`52c7e15367e8306fbca1aa2f8fd7c0a7450ddd8c105e5e912363309778cd3ef6`
MD5	`429ed781bae8e176f3ec94aeacc1468c`
BLAKE2b-256	`57d2e54f24b27b07938b42d04ca5541d23afe545f921cfe03714b058ff88124a`

Algorithm	Hash digest
SHA256	`015d691d78e8634237f6546a708a5708335b5af86891c51ce111b69f40b0790f`
MD5	`7fbdb40508b855d51a199290e09ee9d7`
BLAKE2b-256	`3e5b24547399582ac6ab363c732f9a92a85f98fd1cad4ead2300b043afccdbe5`

Algorithm	Hash digest
SHA256	`75ee1fce0c40845065f8ac2e5d0d3b120a976de3e9d4732a6e99dbaef89afe6c`
MD5	`c08ed6e10a4a34c6df84c16080bacff3`
BLAKE2b-256	`7df4d2f89f0c06539ff6425c6931627dcd362edb620040ac7ab6857edb3053b5`