zipwire

Extract individual files from remote ZIP archives over HTTP - without downloading the whole thing.

A zip wire gets you straight to your destination. This library does the same: it uses HTTP range requests to fetch only the central directory and the specific entries you ask for, skipping everything else. A 10 KB file inside a 2 GB archive? zipwire downloads roughly 10 KB (plus a small overhead for the central directory), not 2 GB.

How it works

ZIP archives store a central directory at the end of the file that lists every entry with its offset and size. zipwire fetches that directory first (a single range request), then makes one additional range request per file you extract. The server must support Range requests (Accept-Ranges: bytes), which most CDNs, object stores, and static file servers do.

Key features

• Selective extraction - download only the files you need, not the entire archive.
• Streaming decompression - read_into decompresses in chunks, keeping memory usage low even for large entries.
• Sync and async - SyncRemoteZip for synchronous code, AsyncRemoteZip with await/async with for asyncio.
• Wheel metadata - SyncRemoteWheel / AsyncRemoteWheel read a Python wheel's .dist-info (METADATA, WHEEL, RECORD) straight from PyPI in a single adaptive tail request, without downloading the wheel.
• ZIP64 - supports archives and entries larger than 4 GiB.
• Pluggable backends - bring your own HTTP library (see below).
• Local files - FileReader / AsyncFileReader open an archive on disk through the exact same API, no HTTP server required.

Installation and backends

The default installation includes the urllib3 backend. To use a different HTTP library, install the matching extra - for example httpx2 gives you both sync and async:

pip install zipwire[httpx2]

Backend	Class	Mode	HTTP	Install extra
urllib3	Urllib3Reader	sync	1.1	(included)
httpx2	Httpx2SyncReader	sync	1.1, 2	httpx2
httpx2	Httpx2AsyncReader	async	1.1, 2	httpx2
requests	RequestsReader	sync	1.1	requests
aiohttp	AiohttpReader	async	1.1	aiohttp

Every HTTP backend accepts an optional pre-configured client or session so you can share connection pools, authentication, and retry configuration.

For archives that already live on the local filesystem, FileReader and AsyncFileReader (in zipwire.backends, no extra dependency) satisfy the same reader protocols - see the local-file example below.

Examples

Read Python wheel metadata without downloading the wheel

A common use case: fetch a wheel's METADATA, WHEEL, or RECORD from PyPI without downloading the (often huge) wheel itself. SyncRemoteWheel and AsyncRemoteWheel parse the wheel URL to locate the .dist-info directory and fetch an adaptive tail, so metadata entries are served from memory without extra HTTP requests:

from zipwire import SyncRemoteWheel
from zipwire.backends import Urllib3Reader

url = "https://files.pythonhosted.org/.../requests-2.32.3-py3-none-any.whl"
with SyncRemoteWheel(Urllib3Reader(url)) as whl:
    print(whl.read(whl.metadata_name).decode())

This optimization relies on the recommended wheel layout of placing .dist-info at the end of the archive; wheels built otherwise still work, falling back to a normal range request per entry.

Sync - list files and read one

from zipwire import SyncRemoteZip
from zipwire.backends import Urllib3Reader

reader = Urllib3Reader("https://archive.example/data.zip")
with SyncRemoteZip(reader) as rz:
    for info in rz.infolist():
        print(f"{info.filename}  {info.file_size} bytes")

    data = rz.read("path/to/file.txt")

Sync - stream a large file to disk

read_into decompresses in chunks so peak memory stays low:

from zipwire import SyncRemoteZip
from zipwire.backends import Urllib3Reader

reader = Urllib3Reader("https://archive.example/large.zip")
with SyncRemoteZip(reader) as rz:
    with open("output.bin", "wb") as f:
        rz.read_into("big-file.bin", f)

Async

import asyncio
from zipwire import AsyncRemoteZip
from zipwire.backends import AiohttpReader

async def main():
    reader = AiohttpReader("https://archive.example/data.zip")
    async with AsyncRemoteZip(reader) as rz:
        data = await rz.read("path/to/file.txt")
        print(data.decode())

asyncio.run(main())

Local archive - same API, no HTTP

FileReader opens an archive from disk through the same interface, accepting a path or a file:// URI. AsyncFileReader is the async counterpart. This lets code that already uses zipwire handle local and remote archives the same way, without special-casing either.

from zipwire import SyncRemoteZip
from zipwire.backends import FileReader

with SyncRemoteZip(FileReader("/path/to/archive.zip")) as rz:
    data = rz.read("path/to/file.txt")

License

Apache-2.0