pvstream

pvstream is a Rust library with python bindings allowing you to efficiently stream download, parse, and filter pageviews from Wikimedia's hourly dumps.

The library can be used from Rust or python. In both languages you can choose between an iterator of parsed objects, made available on the fly as the file is downloaded, or a complete parquet file of parsed and filtered data.

Installation

Rust

Add pvstream to your Cargo.toml:

[dependencies]
pvstream = "0.1"

Or use cargo-add:

cargo add pvstream

Python

Install from PyPI:

pip install pvstream

Building from Source

To build the Python package for your specific hardware:

pip install maturin
git clone https://github.com/vegardege/pvstream
cd pvstream
maturin develop --release

Or build a wheel:

maturin build --release
pip install target/wheels/pvstream-*.whl

Usage

There are four main entry points for this library:

Function	Input	Output
`stream_from_file`	Filename on the local file system	Iterator of parsed row structs
`stream_from_url`	URL of a remotely stored file	Iterator of parsed row structs
`parquet_from_file`	Filename on the local file system	Parquet file of parsed row structs
`parquet_from_url`	URL of a remotely stored file	Parquet file of parsed row structs

[!CAUTION] The _url functions will stream the file directly from Wikimedia's servers. Please be kind to the servers and cache if you plan to read the same file more than once. Consider using a mirror closer to you. You can find mirrors listed on wikimedia.org.

They all accept similar filters. In python, Regex is a str, Vec is a list, u32 is an int:

Filter	Type	Description
`line_regex`	`Option<Regex>`	Regular expression used to filter lines before parsing
`page_title`	`Option<Regex>`	Regular expression used to filter page titles after parsing
`domain_codes`	`Option<Vec<String>>`	List of domain codes to accept
`min_views`	`Option<u32>`	Minimum amount of views needed to be accepted
`max_views`	`Option<u32>`	Maximum amount of views allowed
`languages`	`Option<Vec<String>>`	List of languages to accept
`domains`	`Option<Vec<String>>`	List of domains to accept
`mobile`	`Option<bool>`	If set, filter on whether the row belongs to a mobile site

Learn more about the format from Wikimedia's documentation.

Example (Rust):

use pvstream::filter::FilterBuilder;
use pvstream::stream_from_file;
use std::path::PathBuf;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Path to your pageviews file
    let path = PathBuf::from("pageviews-20240818-080000.gz");

    // View all English mobile sites containing the word 'Rust'
    let filter = FilterBuilder::new()
        .domain_codes(["en.m"])
        .page_title("Rust")
        .build();

    // Stream rows matching the filter
    let rows = stream_from_file(path, &filter)?;

    // Iterate over results
    for row in rows {
        match row {
            Ok(pageview) => println!("{:?}", pageview),
            Err(e) => eprintln!("Error parsing row: {:?}", e),
        }
    }

    Ok(())
}

Example (python):

import pvstream

rows = pvstream.stream_from_file(
    "pageviews-20240818-080000.gz",
    domain_codes=["en.m"],
    page_title="Rust",
)

for row in rows:
    print(row)

pvstream 0.1.0