searching agenda from TRAMP

searching agenda from TRAMP

[JARZz]

[-- Attachment #1: Type: text/plain, Size: 1433 bytes --]

Hello all,

I have a case of a somewhat irregular org mode setup. Here are the givens:

- I have many org files. One for each week of the year, including a couple of generic ones, IE journal.org, routines.org, wiki.org, etc. All in all it's about 120 files or so, all loaded to my org-agenda.

- I've set up Emacs on a mac at work. I do not want my org files stored locally on that machine since it does not belong to me. Instead, I utilized using TRAMP to load the files. I access my remote ssh server with a key and it works OK.

- I need to use agenda searches with specific information, such as user names, machines names, program names etc. in my agenda to find specific cases.

As long as I worked with my files locally, things were OK. TRAMP slows me down a lot, though. I know the problem is not the connection because scp and ssh session works like a breeze. It seems the issues happens with indexing the files.

I don't know what happens behind the scenes exactly, but it takes a good minute or to load my agenda with all the files, and it can take 10-15 minutes to search through them for a specific string. This problem also happens when I used sshfs, which makes me believe more strongly the problem is, again, not with the ssh connection itself, but with the indexing.

Is there a way to seep this up? Or perhaps a workaround you think might work? I want to use the GUI Emacs. What do you think? Also, what causes this?

[-- Attachment #2: Type: text/html, Size: 1627 bytes --]

Re: searching agenda from TRAMP

[Adam Porter]

JARZz <jarss@protonmail.com> writes:

> Hello all, 
>
> I have a case of a somewhat irregular org mode setup. Here are the givens: 
>
> 1 I have many org files. One for each week of the year, including a
> couple of generic ones, IE journal.org, routines.org, wiki.org,
> etc. All in all it's about 120 files or so, all loaded to my
> org-agenda.

120 files is a lot, but the overhead is generally in setting up each Org
buffer, activating org-mode and calling hooks, etc.  Once a buffer is
open, searching them should be fast enough, depending on size, search
type, etc.

> 3 I need to use agenda searches with specific information, such as
> user names, machines names, program names etc. in my agenda to find
> specific cases.

You might give org-ql a try, as its searching is generally much faster,
although it won't solve any TRAMP-related problems.

> As long as I worked with my files locally, things were OK. TRAMP slows
> me down a lot, though. I know the problem is not the connection
> because scp and ssh session works like a breeze. It seems the issues
> happens with indexing the files.

> I don't know what happens behind the scenes exactly, but it takes a
> good minute or to load my agenda with all the files, and it can take
> 10-15 minutes to search through them for a specific string. This
> problem also happens when I used sshfs, which makes me believe more
> strongly the problem is, again, not with the ssh connection itself,
> but with the indexing.

There is no indexing.  :)  (org-ql does implement a per-buffer, per-query
cache for repeated queries, though.  And org-roam does offer a database
that searches some parts of Org files, although I've no idea if it's
compatible with TRAMP.)

> Is there a way to seep this up? Or perhaps a workaround you think
> might work? I want to use the GUI Emacs. What do you think? Also, what
> causes this?

As you probably know, Emacs and Org are full of configuration and
state.  I have little experience with TRAMP, but from what I do have,
I've seen that state and configuration can cause major performance
issues in unexpected ways.  So I'd suggest that you try to reproduce the
problem in a clean, or as minimal as possible, Emacs configuration.  My
with-emacs.sh script would help with that:

https://github.com/alphapapa/with-emacs.sh

If it still happens in a clean Emacs, then you should try using the
latest version of everything: Emacs, Org, and related packages.  (Again,
with-emacs.sh makes it easy to test in separate configs, and something
like Guix can help with running newer Emacs versions.)  If it doesn't
happen in a clean Emacs, then you could try using the bug-hunter package
to bisect your config.

If none of that helps, you'll probably have to dig in to the problem,
use the profiler, etc, to try to find what exactly is taking so long.
This kind of problem usually isn't easy to troubleshoot.  :(

Good luck!

Adam

Re: searching agenda from TRAMP

[Ihor Radchenko]

JARZz <jarss@protonmail.com> writes:

> I don't know what happens behind the scenes exactly, but it takes a good minute or to load my agenda with all the files, and it can take 10-15 minutes to search through them for a specific string. This problem also happens when I used sshfs, which makes me believe more strongly the problem is, again, not with the ssh connection itself, but with the indexing.

Does opening org files manually take a lot of time for you?
Are you using org-id?
Can you track down the problematic function using M-x profiler-start /
profiler-report?

Best,
Ihor

Re: searching agenda from TRAMP

[JARZz]

> Does opening org files manually take a lot of time for you?
>
> Are you using org-id?
>
> Can you track down the problematic function using M-x profiler-start /

These are good questions.
I *do* use org-id. I started doing this more recently. I use a package called super-links. That could be a cause of it?

I can run profiler-start but I'm not sure what I'm looking for. Everything works OK (as far as I can tell), it's only when I want to search something specific it takes forever and a half. But of course, I'll take the advise above and try to run clean Emacs, see what happens then.

Re: searching agenda from TRAMP

[Ihor Radchenko]

JARZz <jarss@protonmail.com> writes:

>> Does opening org files manually take a lot of time for you?
>>
>> Are you using org-id?
>>
>> Can you track down the problematic function using M-x profiler-start /
>
> These are good questions.
> I *do* use org-id. I started doing this more recently. I use a package called super-links. That could be a cause of it?

Can you temporarily set org-id-track-globally to nil and try whatever
commands is slow for you?

> I can run profiler-start but I'm not sure what I'm looking for. Everything works OK (as far as I can tell), it's only when I want to search something specific it takes forever and a half. But of course, I'll take the advise above and try to run clean Emacs, see what happens then.

As a generic advice, you can do profiler-start -> run the command that
you know is slow -> profiler-report -> post the report here.  Profiler
report can be as helpful as debugger backtrace when tracking down
performance issues.

Best,
Ihor