Crate oscar_io

oscar-io

Types and IO (Reader/Writer) for OSCAR Corpus processing and generation.

The crate provides basic abstractions around Corpus items and generic readers/writers useable in OSCAR Corpus files. At some time, it should replace reader implementations in both Ungoliant and oscar-tools.

Features

oscar-io aims to provide readers/writers for numerous types of OSCAR Corpora.

OSCAR v2

• Reader
  • Uncompressed oscar_doc::Reader::new
  • GZipped oscar_doc::Reader::from_gzip
  • Parquet
• Writer
  • Uncompressed oscar_doc::Writer::new
  • GZipped oscar_doc::Writer::new (using a [GzEncoder] reader, from_gzip not yet implemented)
  • Parquet
• SplitReader (Should be unified with SplitReader with split_size: Option<u64>)
  • Uncompressed
  • GZipped
• SplitWriter (Same)
  • Uncompressed
  • GZipped

OSCAR v1.1

• Reader
• Writer
• SplitReader (Should be unified with SplitReader with split_size: Option<u64>)
• SplitWriter (Same)

OSCAR v1

• Reader
• Writer
• SplitReader
• SplitWriter

Modules

common

Common types used in multiple (if not all) different OSCAR Corpus versions.

error

lang

Language enumerations

oscar_doc

OSCAR Schema v2 (OSCAR 22.01) types, readers and writers.