fetchResource loads actual items from the API due to wrong queryParameter

[Sirs0ri]

API Platform version(s) affected: 2.6.8

Description

The API doc parser uses fetch() via the fetchResource() method to load API metadata, and fails to apply the query parameter { itemsperpage: 0 } correctly to limit the request to only metadata and no API items.

fetchResource() tries to set the query parameter via a secondary options parameter, however fetch() doesn't support that - params have to be part of the url instead.

// Current code:
fetchJsonLd(
  resourceUrl,
  Object.assign({ itemsPerPage: 0 }, options)
)

// Fixed code:
fetchJsonLd(
  resourceUrl + "?itemsPerPage=0",
  options
)

• Link to fetchJsonLd
• Link to fetchResource calling fetch()
• fetch() docs on mdn

How to reproduce

Possible Solution

Replace the Object.assign() call with an URL including the query parameter, possibly by building an URL:

const url = new URL(resource.url);
const params = new URLSearchParams(["itemsPerPage", 0]);
url.search = params.toString();

fetchJsonLd(url, options)

Additional Context

Please excuse that my code examples are in plain JS, not TS, since I'm not familiar enough with typescript.

[fzaninotto]

ping @soyuka @alanpoulain this one is a real bummer: in every list page of API Platform Admin, it causes a double fetch to the getList endpoint, one without params for introspection (but still returning data), and the other with the correct params.

[zolex]

ping @soyuka @alanpoulain this one is a real bummer: in every list page of API Platform Admin, it causes a double fetch to the getList endpoint, one without params for introspection (but still returning data), and the other with the correct params.

actually it's not always just two. i have scenarios where it does this more than 10 times before and after the actual fetch with the query params... see example here: https://github.com/api-platform/admin/issues/627

[soyuka]

I'm unsure how we should fix this on an RDF point of view, ie while respecting hydra's specification. On discovery we don't know the pagination parameters, also using HTTP cache should definitely mitigate these as the discovery queries should already be cached by your browser.

[zolex]

I'm unsure how we should fix this on an RDF point of view, ie while respecting hydra's specification. On discovery we don't know the pagination parameters, also using HTTP cache should definitely mitigate these as the discovery queries should already be cached by your browser.

what about adding an option to allow hooking into an actual collection request from the application, the doc-parser is used in?