git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH] fsck: print progress

From
Jeff King <peff@peff.net>
Date
Nov 3, 2011, 19:51 UTC
Message-ID
<20111103195147.GA21318@sigill.intra.peff.net>
In-Reply-To
<7vd3d9f0u8.fsf@alter.siamese.dyndns.org>
On Thu, Nov 03, 2011 at 12:43:59PM -0700, Junio C Hamano wrote:
Show 19 quoted lines
> > But I almost wonder if it is worth factoring out the concept of a
> > "progress group", where you would call it like:
> >
> >   progress = progress_group_start("Checking objects in pack", nr_packs);
> >   for (p = packed_git; p; p = p->next) {
> >           progress_group_next(progress, p->num_objects);
> >           for (j = 0; j < num; j++) {
> >                   fsck_sha1(p, j);
> >                   display_progress(progress, j+1);
> >           }
> >   }
> >   stop_progress(&progress);
> 
> Hmm, once you do this kind of thing I would expect two (or more) progress
> bars to be shown, something like:
> 
> 	$ git fsck --progress
> 	checking packs: 3 of 7 (42% done)
>         checking objects in pack: 12405 of 332340 (4% done)

I don't think we can do multiple lines portably, though, because the progress code just uses "\r" to return to the beginning of the line.

However, I do think it's a nice design even if the output is the same right now, because the calling code is specifying more clearly to the progress code what it actually means. Which means it is easier to tweak the progress code later to make a prettier representation of that meaning.

One downside (as with all abstractions :) ), is that it's hard to deviate from the defaults. With the above, how would you specify if it was a group of throughput measurements, not counts? Or that you wanted to use start_progress_delay instead of start_progress on each one?

Or hey, if you really want to get crazy, why can't we check each pack in parallel on a different CPU, and have all of the progress meters displayed and moving simultaneously? :)

The last one is obviously a bit tongue in cheek. But it does raise an interesting point: is seeing the per-pack meter actually useful (whether parallel or not)? The user just wants to know that progress is being made, and when it is done. And maybe that argues for just summing the size of each pack and showing a single progress meter per task.

-Peff
Previous: Junio C HamanoNext: Junio C Hamano
Message 10 of 21 in “long fsck time”
  1. Nguyen Thai Ngoc DuyNov 2, 2011
  2. Nguyen Thai Ngoc DuyNov 2, 2011
  3. Jeff KingNov 2, 2011
  4. Nguyen Thai Ngoc DuyNov 3, 2011
  5. fsck: print progressNguyễn Thái Ngọc Duy, Nov 3, 2011
  6. Jeff KingNov 3, 2011
  7. fsck: print progressNguyễn Thái Ngọc Duy, Nov 3, 2011
  8. Jeff KingNov 3, 2011
  9. Junio C HamanoNov 3, 2011
  10. Jeff KingNov 3, 2011
  11. Junio C HamanoNov 3, 2011
  12. Jeff KingNov 3, 2011
  13. Junio C HamanoNov 3, 2011
  14. Jeff KingNov 3, 2011
  15. fsck: print progressNguyễn Thái Ngọc Duy, Nov 4, 2011
  16. Stephen BoydNov 5, 2011
  17. Nguyen Thai Ngoc DuyNov 5, 2011
  18. Stephen BoydNov 5, 2011
  19. Junio C HamanoNov 5, 2011
  20. Nguyen Thai Ngoc DuyNov 6, 2011
  21. Junio C HamanoNov 6, 2011

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.