|
From: | Dmitry Gutov |
Subject: | Re: A project-files implementation for Git projects |
Date: | Fri, 4 Oct 2019 01:49:23 +0300 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.8.0 |
On 03.10.2019 20:15, Tassilo Horn wrote:
I just tried with the gcc repository with its about 100000 files. Here it was about equal with ~30secs for both git and find. The git ls-files invocation on the command line with output piped to /dev/null is done in one tenth of a second. Oh, when dropping the `expand-file-name' call we're doing on every file, it's ten times faster (~3secs).
Nice observation, thanks. It's a significant per-item cost, so no surprise projects with lots of files are (were) doing worse than 'find'.
I've changed it to 'concat' now which isn't free, but much faster.I wish someone did some benchmarking with Hg projects. mozilla-central (~200000 files) is still slower on my machine with Hg than with 'find'.
We still don't support regexp rules with 'find', but that's probably doesn't affect the speed in this example, and it's a reason to prefer Hg for listing.
[Prev in Thread] | Current Thread | [Next in Thread] |