@Pchelolo @mobrovac - some background - we need to do better at signaling when an extract is not available. This might be because the page is in a namespace where extracts are not possible or there is no lead section to generate one from. e.g. https://en.wikipedia.beta.wmflabs.org/api/rest_v1/page/summary/Jon%20example

It feels like 404ing is the right thing to do here - but what do you think?

It feels like 404ing is the right thing to do here - but what do you think?

Hm.. We can 404 if the extract and Wikidata description are empty, that' easy. Are we sure there's no corner case where the extract is not available while some other info might be available like a page image or coordinate? Hard to imaging such a case though..

Right now it's a summary endpoint so I think it will only ever contain summaries. The idea would be to to error if nothing is available.

extract_html is empty when rendered has no textContent

Can you give me an example when the extract_text is not empty while the textContent is? How can that happen?

Created a PR for it https://github.com/wikimedia/restbase/pull/831 but I'm not too sure if empty text extract is a good enough approximation for empty textContent

So we have already had T167045: exsentences incorrectly returns empty text extract where not having an extract crashes the Android app if the field is not present. That means that omitting the field is not an option. Throwing a 404 if the field is not present is a bit contentious I think: it is true that the extract is the centrepiece of the summary end point response, but there are other bits of information that conceptually make up what we think of a page's summary is. IMHO, returning an empty extract is the correct way to go here (the client can always display nothing). Returning a 404 should be done when either (a) the page doesn't exist; or (b) no information is truly available for that page.

In T168418#3368924, @Pchelolo wrote:

Can you give me an example when the extract_text is not empty while the textContent is? How can that happen?

In the case of an HTML extract, this could be something like '<p><b>(Blank.)</b></p>' and that would be transformed to '<p><b></b></p>' after the parenthetical statement is stripped. I'm not sure this could happen with a plain text extract.

In T168418#3374752, @phuedx wrote:

I'm not sure this could happen with a plain text extract.

Because we have recently switched to asking for the HTML and then stripping the tags (cf T165017 for context).