REST-like CRUD+ HTTP API for index patterns

[mattkime]

Create a REST API for index patterns—a wrapper around the Index Patterns service.

• CRUD
  • Create — POST /api/index_patterns/index_pattern
  • Read — GET /api/index_patterns/index_pattern/{id}
  • Delete — DELETE /api/index_patterns/index_pattern/{id}
• Update portion of an index pattern — POST /api/index_patterns/index_pattern/{id}
  • title
  • timeFieldName
  • intervalName
  • fields
    • Optionally refresh fields.
  • sourceFilters
  • fieldFormatMap
  • type
  • typeMeta
• Fields API
  • Create a new field — POST /api/index_patterns/index_pattern/{id}/field
  • Read a new field — GET /api/index_patterns/index_pattern/{id}/field/{name}
  • Update a an existing field — POST /api/index_patterns/index_pattern/{id}/field/{name}
  • Upsert a field — PUT /api/index_patterns/index_pattern/{id}/field
  • Remove a field — DELETE /api/index_patterns/index_pattern/{id}/field/{name}

[elasticmachine]

Pinging @elastic/kibana-app-arch (Team:AppArch)

[samtriolo]

A couple use cases for myself -

1. creating an index pattern after I’ve created a new index - all I typically need to do is select the index name and default time series, if applicable, and it’s gtg.

2. modify an existing index pattern - remove or update a field in an existing index pattern.

[walterra]

This could be quite useful if you're using other parts of ES via APIs programmatically and in the same workflow you'd want to make indices accessible for Kibana users without manually creating an index pattern yourself.

For example, a Python developer might use Machine Learning analytics. The results will be written to a destination index. When using the ML plugin's Kibana UI you can opt in to create an index pattern, but if you're just using the ES API in Python that option is not available. So you'd need to go to Kibana and manually create an index pattern so other users can view the results for example in Kibana Discover.

Here's a repository with an exemplary workflow on how data scientists can use Python/Jupyter notebooks in combination with Kibana: https://github.com/walterra/jupyter2kibana — a REST API for index patterns would be a great addition!

Even our own Python library eland (https://github.com/elastic/eland) could make use of that API under the hood! https://github.com/elastic/eland

[samtriolo]

@walterra explained it much better than I did, but that's basically the use case I was trying to outline in my comment. I create indices programmatically all the time, and yet there's always the manual step of having to create the index pattern. For me, it's only a couple things I specify each time, eg the time series, so it seems it would be relatively straightforward to build a programmatic interface for.