csv-partial-cache

Csv index with partially cached columns

It meant to allow a performant access to immutable csv data without importing it to database. The idea is to keep line offsets with frequently used columns in memory, while accessing the full line following the offset.

Usage

Let's say we have a table of http statuses:

And we'd like to cache short code and name columns.

use csv_partial_cache::{self, CsvPartialCache, FromLineOffset};

/// Full table row representation
#[derive(serde::Deserialize)]
struct FullRecord {
    code: u16,
    name: String,
    description: String,
}

/// Columns `code` and `name` are essential and small, so we'd like to keep them and memory
struct PartialRecord {
    code: u16,
    name: String,
    // The file offset to access the full record data
    offset: u16,
}

// We should tell a little more about the cached representation
impl FromLineOffset for PartialRecord {
    /// type of file offsets used instead of default `u64` to spare some memory
    type Offset = u16;

    /// Pointer to the offset field
    fn offset(&self) -> Self::Offset {
        self.offset
    }

    /// Constructor from the line and it's offset
    fn from_line_offset(line: &str, offset: Self::Offset) -> csv_partial_cache::Result<Self> {
        let (code, name) = csv_line::from_str(line)?;
        Ok(Self { code, name, offset })
    }
}

tokio_test::block_on(async {
    // Now we can load the index
    let cache = CsvPartialCache::<PartialRecord>::new("tests/status_codes.csv").unwrap();

    // Getting a partial data cached in memory
    let partial = cache.find(&100, |x| x.code).unwrap();
    assert_eq!(partial.name, "Continue");

    /// Loading full data from the csv file
    let full: FullRecord = cache.full_record(&partial).await.unwrap();
    assert_eq!(full.code, partial.code);
})

Contributing

Please run .pre-commit.sh before submitting a pull request to ensure all checks pass.

License

This project is licensed under the MIT license.