group search results by source package #1

Closed
mbiebl opened this Issue Mar 26, 2013 · 4 comments

Comments

Projects
None yet
2 participants
@mbiebl

mbiebl commented Mar 26, 2013

seems I found the bug tracker for codesearch :-)

codesearch is already awesome, what would make it even more awesome to do archive wide greps and mass bug filings, is the ability to group the search results by source package and only show a summary per source package which can be expanded via a +

@stapelberg

This comment has been minimized.

Show comment
Hide comment
@stapelberg

stapelberg Mar 26, 2013

Contributor

Just FYI: Complexity-wise, this is ≈ O(full-search), because nearly every file needs to be looked at.

In order to implement that feature, an entirely new code path needs to be implemented which uses short-circuits on source package level.

Depending on how long such a search query typically takes, it might be necessary to add some user interface to perform a long-running query. That is, the client should have a way to abort the query and see progress nearly as it happens.

Contributor

stapelberg commented Mar 26, 2013

Just FYI: Complexity-wise, this is ≈ O(full-search), because nearly every file needs to be looked at.

In order to implement that feature, an entirely new code path needs to be implemented which uses short-circuits on source package level.

Depending on how long such a search query typically takes, it might be necessary to add some user interface to perform a long-running query. That is, the client should have a way to abort the query and see progress nearly as it happens.

@stapelberg

This comment has been minimized.

Show comment
Hide comment
@stapelberg

stapelberg Jul 3, 2013

Contributor

I had an idea on how to make “batch” queries work:

Each service file should include an instance specifier (e.g. dcs-web@batch0.service) and run the entire stack, but with lower I/O and CPU priority, so that normal queries are not impacted. The user can then reliably cancel queries (by bringing down the stack) and doesn’t impact normal usage.

It should be verified that mmaping the same (index) file in two separate processes accesses the same memory in the kernel.

Contributor

stapelberg commented Jul 3, 2013

I had an idea on how to make “batch” queries work:

Each service file should include an instance specifier (e.g. dcs-web@batch0.service) and run the entire stack, but with lower I/O and CPU priority, so that normal queries are not impacted. The user can then reliably cancel queries (by bringing down the stack) and doesn’t impact normal usage.

It should be verified that mmaping the same (index) file in two separate processes accesses the same memory in the kernel.

@stapelberg

This comment has been minimized.

Show comment
Hide comment
@stapelberg

stapelberg Jul 27, 2013

Contributor

Commit 23da966 lays the foundation for my most recent comment.

What still needs to be done is build a UI around the batch queries.

Contributor

stapelberg commented Jul 27, 2013

Commit 23da966 lays the foundation for my most recent comment.

What still needs to be done is build a UI around the batch queries.

@stapelberg

This comment has been minimized.

Show comment
Hide comment
@stapelberg

stapelberg Dec 8, 2014

Contributor

With the launch of Debian Code Search Instant, this is fixed (too many commits to mention them all in this issue).

Contributor

stapelberg commented Dec 8, 2014

With the launch of Debian Code Search Instant, this is fixed (too many commits to mention them all in this issue).

@stapelberg stapelberg closed this Dec 8, 2014

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment