bookdata/cli/cluster/author_gender/
clusters.rs

//! Read cluster information.
use std::collections::HashMap;
use std::convert::identity;
use std::path::Path;

use super::authors::AuthorTable;
use crate::arrow::scan_parquet_file;
use crate::gender::*;
use crate::prelude::*;
use crate::util::logging::item_progress;
use anyhow::Result;
use parquet_derive::ParquetRecordReader;
use polars::prelude::*;

/// Record for storing a cluster's gender statistics while aggregating.
#[derive(Debug, Default)]
pub struct ClusterStats {
    pub n_book_authors: u32,
    pub n_author_recs: u32,
    pub genders: GenderBag,
}

/// Row struct for reading cluster author names.
#[derive(Debug, ParquetRecordReader)]
struct ClusterAuthor {
    cluster: i32,
    author_name: String,
}

pub type ClusterTable = HashMap<i32, ClusterStats>;

/// Read cluster author names and resolve them to gender information.
pub fn read_resolve(path: &Path, authors: &AuthorTable) -> Result<ClusterTable> {
    let timer = Timer::new();
    info!("reading cluster authors from {}", path.display());
    let iter = scan_parquet_file(path)?;

    let pb = item_progress(iter.remaining() as u64, "authors");

    let mut table = ClusterTable::new();

    for row in pb.wrap_iter(iter) {
        let row: ClusterAuthor = row?;
        let rec = table.entry(row.cluster).or_default();
        rec.n_book_authors += 1;
        if let Some(info) = authors.get(row.author_name.as_str()) {
            rec.n_author_recs += info.n_author_recs;
            rec.genders.merge_from(&info.genders);
        }
    }

    info!(
        "scanned genders for {} clusters in {}",
        table.len(),
        timer.human_elapsed()
    );

    Ok(table)
}

/// Read the full list of cluster IDs.
pub fn all_clusters<P: AsRef<Path>>(path: P) -> Result<Vec<i32>> {
    info!("reading cluster IDs from {}", path.as_ref().display());
    let path = path
        .as_ref()
        .to_str()
        .map(|s| s.to_string())
        .ok_or(anyhow!("invalid unicode path"))?;
    let df = LazyFrame::scan_parquet(path, Default::default())?;
    let df = df.select([col("cluster")]);
    let clusters = df.collect()?;
    let ids = clusters.column("cluster")?.i32()?;

    info!("found {} cluster IDs", ids.len());

    Ok(ids.into_iter().filter_map(identity).collect())
}
bookdata/cli/cluster/author_gender/clusters.rs

bookdata/cli/cluster/author_gender/
clusters.rs
