libzim
module allows you to read and write ZIM
files in Python. It provides a shallow python
interface on top of the C++ libzim
library.
It is primarily used in openZIM scrapers like sotoki
or youtube2zim
.
pip install libzim
Our PyPI wheels bundle a recent release of the C++ libzim and are available for the following platforms:
x86_64
and arm64
x86_64
, armhf
and aarch64
x86_64
and aarch64
x64
Wheels are available for CPython only (but can be built for Pypy).
Users on other platforms can install the source distribution (see Building below).
git clone git@github.com:openzim/python-libzim.git && cd python-libzim
# hatch run test:coverage
See CONTRIBUTING.md for additional details then Open a ticket or submit a Pull Request on Github 🤗!
from libzim.reader import Archive
from libzim.search import Query, Searcher
from libzim.suggestion import SuggestionSearcher
zim = Archive("test.zim")
print(f"Main entry is at {zim.main_entry.get_item().path}")
entry = zim.get_entry_by_path("home/fr")
print(f"Entry {entry.title} at {entry.path} is {entry.get_item().size}b.")
print(bytes(entry.get_item().content).decode("UTF-8"))
# searching using full-text index
search_string = "Welcome"
query = Query().set_query(search_string)
searcher = Searcher(zim)
search = searcher.search(query)
search_count = search.getEstimatedMatches()
print(f"there are {search_count} matches for {search_string}")
print(list(search.getResults(0, search_count)))
# accessing suggestions
search_string = "kiwix"
suggestion_searcher = SuggestionSearcher(zim)
suggestion = suggestion_searcher.suggest(search_string)
suggestion_count = suggestion.getEstimatedMatches()
print(f"there are {suggestion_count} matches for {search_string}")
print(list(suggestion.getResults(0, suggestion_count)))
from libzim.writer import Creator, Item, StringProvider, FileProvider, Hint
class MyItem(Item):
def __init__(self, title, path, content = "", fpath = None):
super().__init__()
self.path = path
self.title = title
self.content = content
self.fpath = fpath
def get_path(self):
return self.path
def get_title(self):
return self.title
def get_mimetype(self):
return "text/html"
def get_contentprovider(self):
if self.fpath is not None:
return FileProvider(self.fpath)
return StringProvider(self.content)
def get_hints(self):
return {Hint.FRONT_ARTICLE: True}
content = """<html><head><meta charset="UTF-8"><title>Web Page Title</title></head>
<body><h1>Welcome to this ZIM</h1><p>Kiwix</p></body></html>"""
item = MyItem("Hello Kiwix", "home", content)
item2 = MyItem("Bonjour Kiwix", "home/fr", None, "home-fr.html")
with Creator("test.zim").config_indexing(True, "eng") as creator:
creator.set_mainpath("home")
creator.add_item(item)
creator.add_item(item2)
illustration = pathlib.Path("icon48x48.png").read_bytes()
creator.add_illustration(48, illustration)
for name, value in {
"creator": "python-libzim",
"description": "Created in python",
"name": "my-zim",
"publisher": "You",
"title": "Test ZIM",
"language": "eng",
"date": "2024-06-30"
}.items():
creator.add_metadata(name.title(), value)
The reading part of the libzim is most of the time thread safe. Searching and creating part are not. libzim documentation
python-libzim
disables the GIL on most of C++ libzim calls. You must prevent concurrent access yourself. This is easily done by wrapping all creator calls with a threading.Lock()
lock = threading.Lock()
with Creator("test.zim") as creator:
# Thread #1
with lock:
creator.add_item(item1)
# Thread #2
with lock:
creator.add_item(item2)
libzim
being a binary extension, there is no Python source to provide types information. We provide them as type stub files. When using pyright
, you would normally receive a warning when importing from libzim
as there could be discrepencies between actual sources and the (manually crafted) stub files.
You can disable the warning via reportMissingModuleSource = "none"
.
libzim
package building offers different behaviors via environment variables
Variable | Example | Use case |
---|---|---|
LIBZIM_DL_VERSION |
8.1.1 or 2023-04-14 |
Specify the C++ libzim binary version to download and bundle. Either a release version string or a date, in which case it downloads a nightly |
USE_SYSTEM_LIBZIM |
1 |
Uses LDFLAG and CFLAGS to find the libzim to link against. Resulting wheel won't bundle C++ libzim. |
DONT_DOWNLOAD_LIBZIM |
1 |
Disable downloading of C++ libzim. Place headers in include/ and libzim dylib/so in libzim/ if no using system libzim. It will be bundled in wheel. |
PROFILE |
0 |
Enable profile tracing in Cython extension. Required for Cython code coverage reporting. |
SIGN_APPLE |
1 |
Set to sign and notarize the extension for macOS. Requires following informations |
APPLE_SIGNING_IDENTITY |
Developer ID Application: OrgName (ID) |
Required for signing on macOS |
APPLE_SIGNING_KEYCHAIN_PATH |
/tmp/build.keychain |
Path to the Keychain containing the certificate to sign for macOS with |
APPLE_SIGNING_KEYCHAIN_PROFILE |
build |
Name of the profile in the specified Keychain |
On Windows, built wheels needs to be fixed post-build to move the bundled DLLs (libzim and libicu) next to the wrapper (Windows does not support runtime path).
After building you wheel, run
python setup.py repair_win_wheel --wheel=dist/xxx.whl --destdir wheels\
Similarily, if you install as editable (pip install -e .
), you need to place those DLLs at the root
of the repo.
Move-Item -Force -Path .\libzim\*.dll -Destination .\
python3 -m build
# using system-installed C++ libzim
brew install libzim # macOS
apt-get install libzim-devel # debian
dnf install libzim-dev # fedora
USE_SYSTEM_LIBZIM=1 python3 -m build --wheel
# using a specific C++ libzim
USE_SYSTEM_LIBZIM=1 \
CFLAGS="-I/usr/local/include" \
LDFLAGS="-L/usr/local/lib"
DYLD_LIBRARY_PATH="/usr/local/lib" \
LD_LIBRARY_PATH="/usr/local/lib" \
python3 -m build --wheel
On platforms for which there is no official binary available, you'd have to compile C++ libzim from source first then either use DONT_DOWNLOAD_LIBZIM
or USE_SYSTEM_LIBZIM
.