Only 100 results?

[lsegal]

I am noticing that most searches are only yielding 5 pages of results. Is this just a side-effect of CodeSearch being in preview mode, or is the expectation that each search will only show the first 100 results?

As someone who has been using GitHub search for total counts, I am extremely interested to know if codesearch will be usable for determining total occurrences of terms across (unique) repositories. The 5 page limit with no indication of total counts makes it extremely difficult to use Code Search in its current form.

The first screenshot below looks way more useful to me than the 2nd:

old search:

new search:

[anotherlab]

We have hundreds of repos that are on .NET 5. As part of our migration to .NET 6, it would be useful to be able to query for .csproj files that are .NET 5 and see all of them so that we can plan on our migrations.

[colinwm]

We are working on an API and intend to allow exhaustive search via the API without the 100 result limit.

As for estimating result set size, we've heard this feedback before and are working on it!

[falcucci]

they downgrade the search and we need to live like that now.

just give the option to rollback it at least.

i am shocked

[beocaca]

Any update please...?

[StashOfCode]

@colinwm It's been two years lol "We are working..." my ass XD

[cobavcova]

any update collin? please give back old code search

[NagaTheDon]

Any update on this?

[adonovan]

I understand the need for limits, but 100 seems very small. Large-scale refactoring and clean-up operations within an organization often touch hundreds of files or call sites. For larger result sets I'd be happy with just a list of affected files, or even repositories, so that I can use the search results as roots and do the rest in my scripts. Increasing this limit by a factor of 10 would be a welcome step.

Regardless of the limit, it would also be very useful to be able to download the result set in a simple grep-like format ("file:line: text"). A dense result set would save the need to scrape filenames out of 5 or more HTML pages of results. I know that a 'gh search' tool is coming and will provide this feature, but it would be useful in the web interface too.

Furthermore, showing complete results within an organization's scope seems like an important feature for clean-up tasks, where you need to know that you've found all matches.

[stephdotnet]

Absolute yes for private repos in organization. This is my use case and wanted to use GH search in this scenario. Didn't since of this limitation 🤷

[catkins]

I'm running into this as well. We're trying to build some statistics across our org on adoption of a few specific internal libraries, but the limit of 100 results is meaning that I can't easily find my answers.

I'd be happy enough with the ability to paginate through all of them with a cursor and a lower rate limit as long as I could retrieve all of the results.

[SeifESlimene]

So is there any solution to the 100 results limit?

[Ponjimon]

The only solution right now is to narrow down your search. Unfortunately.

[falcucci]

they did rollout the new search and don't even offer a fallback solution.

that's really bad.

[falcucci]

unbelievable

[rhanton]

Ugh, this is crazy.

[adonovan]

unbelievable

Ugh, this is crazy.

Regardless of how much frustration you might feel, these are not constructive ways to vent it (and if you made these comments on my issue tracker, I would feel less, not more, inclined to help you). Bear in mind that software maintainers always have dozens of things competing for their attention. If this issue is important to you, argue the case.

[falcucci]

Regardless of how much frustration you might feel, these are not constructive ways to vent it (and if you made these comments on my issue tracker, I would feel less, not more, inclined to help you). Bear in mind that software maintainers always have dozens of things competing for their attention. If this issue is important to you, argue the case.

seems that you didn't get that this has nothing with some specific developer/field. They are still here trying to help creating issues and using the platform.

It's just people tired about the mistakes the company did, messing one of the most used feature by them.

It's just a genuine expression about what happened here and how github is handling it.

The issue itself has enough detail about what they miss.

Unbelievable.

[StashOfCode]

It's just Microshit trying to prevent LLMs other than their chatgpt getting trained on open source code. Github is done.

[theharshpat]

hey what's up, any update on this?

[sgf]

I really can't understand Microsoft's behavior. You don't reduce costs through technology, but rely on castration of functions and degrading the experience. This is really puzzling.

[felix-d1strict]

@colinwm Sorry to ping you, but do you have any update on this?

[llakala]

Someone said to argue my case for this issue above, so here goes.

Github search is really useful for finding real-world results that I can use as references for my code. Regardless of some of its flaws, it's an extraordinary tool. But its power is limited by the fact that it's unable to show more than 100 results. I wouldn't even complain that much if the limit was increased a bit for now as a temporary solution (I'd be happy with 250). I know that API development takes time, and that there are other priorities. But this is affecting a lot of users and would improve the experience of a lot of people.

The solution doesn't need to be perfect immediately, but seeing any sort of improvement to this, even a minor one, would be much appreciated.

[theharshpat]

They don't want AI companies scrapping the data easily. But 100 is just way too less, it should be increased to 250 results at the very least.

[jpivarski]

The reason can't be to avoid scraping for AI, since all of the data is also available at https://www.gharchive.org/ (complete public metadata logs) and by git-cloning the repos directly (complete public git data). Both of these are in a more bulk-digestable form—good for AI training—than the search tool. The search tool is just for humans.

By the way, for my use-case of finding all users of my library, I now scrape the Insights/Dependency Graph (only available as HTML, as far as I can tell), then git-clone all of those repos and do an offline text search. GitHub is (thankfully!) very generous in how quickly it can deliver git data.

But this only underscores the point that the search tool is for humans. I wouldn't want to git-clone a repo just to search it for a keyword.

[sgf]

I hold the opposite view. I can only accept progress and cannot accept regression.

This should be obvious, I don't know why regression occurs. Is the existence of developers a reverse promotion effect? ​​If the developer is not capable enough, it is recommended to replace him. Don't stand there digging a hole without doing anything.

I don't know what problems GitHub has encountered since it was acquired by Microsoft. But it is obvious that many functions such as dashboard-feed have been reverse optimized. I seriously doubt whether Microsoft Group's management ability is insufficient.

Or maybe Microsoft is simply not good at creating and operating a large social network. Looking at Microsoft's history, I'm not making this up, but there are clues to follow.

Countless feedback from the community was created, but GitHub insisted on its own way, and these reverse optimizations had no precedent for improvement.

[theharshpat]

Please make it at least 250 results

[NagaTheDon]

Any update on this?? Why are we only allowed to see only 100 results? What happened to the old search?

[tbadalov]

Can't see beyond 5 pages of results. This is too small for the number of files actually found for a query. Hopefully this will get fixed soon.

[Floccinaucinihilipilification11]

For the past couple of weeks I suddenly started getting up to 100 pages of results (yes, up to 1000 results). I assume this is true for everyone?

[jpivarski]

Still just 5 pages (100 results) for me.

[Xaypanya]

Still 5 pages here

[Floccinaucinihilipilification11]

I get 100 pages with 10 results per page. Maybe it is slowly rolling out.