omniparser

Omniparser is a native Golang ETL parser that ingests input data of various formats (CSV, txt, fixed length/width, XML, EDI/X12/EDIFACT, JSON, and custom formats) in streaming fashion and transforms data into desired JSON output based on a schema written in JSON.

Golang Version: 1.14

Documentation

Docs:

• Getting Started: a tutorial for writing your first omniparser schema.
• IDR: in-memory data representation of ingested data for omniparser.
• XPath Based Record Filtering and Data Extraction: xpath queries are essential to omniparser schema writing. Learn the concept and tricks in depth.
• All About Transforms: everything about transform_declarations.
• Use of custom_func, Specially javascript: An in depth look of how custom_func is used, specially the all mighty javascript (and javascript_with_context).
• CSV Schema in Depth: everything about schemas for CSV input.
• Fixed-Length Schema in Depth: everything about schemas for fixed-length (e.g. TXT) input
• JSON/XML Schema in Depth: everything about schemas for JSON or XML input.
• EDI Schema in Depth: everything about schemas for EDI input.
• Programmability: Advanced techniques for using omniparser (or some of its components) in your code.

References:

• Custom Functions: a complete reference of all built-in custom functions.

Examples:

• CSV Examples
• Fixed-Length Examples
• JSON Examples
• XML Examples.
• EDI Examples.
• Custom File Format
• Custom Parse

In the example folders above you will find pairs of input files and their schema files. Then in the .snapshots sub directory, you'll find their corresponding output files.

Online Playground

Use https://omniparser.herokuapp.com/ (may need to wait for a few seconds for heroku instance to wake up) for trying out schemas and inputs, yours or existing samples, to see how ingestion and transform work.

Why

• No good ETL transform/parser library exists in Golang.
• Even looking into Java and other languages, choices aren't many and all have limitations:
  • Smooks is dead, plus its EDI parsing/transform is too heavyweight, needing code-gen.
  • BeanIO can't deal with EDI input.
  • Jolt can't deal with anything other than JSON input.
  • JSONata still only JSON -> JSON transform.
• Many of the parsers/transforms don't support streaming read, loading entire input into memory - not acceptable in some situations.

Requirements

• Golang 1.14

Recent Major Feature Additions/Changes

• Added fixed-length file format support in omniv21 handler.
• Added EDI file format support in omniv21 handler.
• Major restructure/refactoring
  • Upgrade omni schema version to omni.2.1 due a number of incompatible schema changes:
    • 'result_type' -> 'type'
    • 'ignore_error_and_return_empty_str -> 'ignore_error'
    • 'keep_leading_trailing_space' -> 'no_trim'
  • Changed how we handle custom functions: previously we always use strings as in param type as well as result param type. Not anymore, all types are supported for custom function in and out params.
  • Changed the way how we package custom functions for extensions: previously we collect custom functions from all extensions and then pass all of them to the extension that is used; This feels weird, now changed to only the custom functions included in a particular extension are used in that extension.
  • Deprecated/removed most of the custom functions in favor of using 'javascript'.
  • A number of package renaming.
• Added CSV file format support in omniv2 handler.
• Introduced IDR node cache for allocation recycling.
• Introduced IDR for in-memory data representation.
• Added trie based high performance times.SmartParse.
• Command line interface (one-off transform cmd or long-running http server mode).
• javascript engine integration as a custom_func.
• JSON stream parser.
• Extensibility:
  • Ability to provide custom functions.
  • Ability to provide custom schema handler.
  • Ability to customize the built-in omniv2 schema handler's parsing code.
  • Ability to provide a new file format support to built-in omniv2 schema handler.

Footnotes

• omniparser is a collaboration effort of jf-tech,Simon and Steven.