gupy-io / searchops Goto Github PK

Commit 15bf22b introduced a performance issue.

Ajv 7.x does not serialize schemas before caching compilation results, thus the following code became dangerous:

export function getValidatorForMapping(
  mappings: ObjectMapping
): ValidateFunction<unknown> {
  return ajv.compile(translateObjectMapping(mappings));
} 

After removing Winston as a hard dependency, the module now may cause compilation errors if the winston peer dependency is not installed.

error TS2307: Cannot find module 'winston' or its corresponding type declarations.

1 import type { Logger } from "winston";

It was important to enable users to install directly from GitHub, but now that the package is published, there's no reason to keep it this way. It removes the need to check for transpilation changes.

The most obvious choice is to use fast-json-stable-serialize, which is exactly what Ajv 6 did.
Even better than caching by schemas like AJV would do, it would be better to cache by mappings.

Solves #163

A few methods on the SearchService class have been coupled to a particular mapping, which infects the class abstraction with a particular use case.

We should expel such code (specially getshould).

Implement the referenced ADR:
https://gupy-io.github.io/searchops/adrlogs/adr/20210203-optional-document-schema-validation

Tasks:

• Cache JSON schemas built from index mappings #165
• Make usage of schema validation opt-in #232

It serves no real purpose, and it requires to manually update the package-lock every time the package changes in order to pass integrity checks.

A singleton client should not be enforced for every user of this library.

If this was commonplace, I wouldn't have to copy over Winston declarations to solve #37. I think this should be come good practice for typescript libraries now that we have import type.

Saw this on the last ElasticON, would be cool to centralize some of those.

● Index Migrations › Updating dynamic index settings › given the index was created › when the manager performs a migration › then the index should exist

    ResponseError: invalid_alias_name_exception

https://github.com/gupy-io/searchops/runs/1936859362#step:6:105