[WIP] remove call to external ls when loading chruby

[kbrock]

before

/Users/kbrock/dotfiles/bashrc.d/chruby.sh

real	0m0.004s
user	0m0.001s
sys	0m0.001s

after

/Users/kbrock/dotfiles/bashrc.d/chruby.sh

real	0m0.000s
user	0m0.000s
sys	0m0.000s

why?

I was trying to figure out ways to make nvm.sh load faster on my system.
chruby loads so much faster, and when I looked to see why, I was distracted by the call to ls -A "$dir". It was using a builtin to list the files in the directory, but an external command to detect if there were files in the directory.

So yea, I converted the already quick method to made it even faster.

• I tested with my files in ~/.rubies, and a non-existent /opt/rubies.
• I did not test this with a directory with a space in the name, but didn't see how that could be possible other than having a PREFIX with a space in the name.
• I did not test with rubies that start with a period, but it looked like the original script did not handle that case either. And it didn't seem valid.

This is yak shaving, and you probably can think up a better implementation. But it seems to work in zsh and bash.

If you prefer not to introduce change (because change always introduces risk), then that is cool. Just close.
I just saw this change and wanted to share.

As you will note, nvm is noticeably slower and the reason for my original tangent:

/Users/kbrock/dotfiles/bashrc.d/node.bash

real	0m0.177s
user	0m0.033s
sys	0m0.089s

Thanks again for such a great tool. And it already performs heads and shoulders better than other tools out there. (ruby or other)

refs:

• https://github.com/kbrock/dotfiles/blob/master/bashrc.d/chruby.sh
• https://github.com/kbrock/dotfiles/blob/master/bashrc.d/node.bash (since changed to speed this up)

[kbrock]

WIP: sorry, in zsh it worked for non-existing director, or a directory with rubies, but not for a directory that was empty

and zsh is complicated due to the default of blowing up when there is a bad glob

running setopt nonomatch will get around this, but this started to get too complicated.
Since it is sourced, modifying that value in the parent shell is not so good.

[postmodern]

@kbrock yep, we ended up using ls -A to filter out non-existent directories or empty directories, and to behave consistently even if the user has enabled some weird glob-option mode.

In the 1.0.0 branch, I changed chruby to lazy enumerate the directories on each call with an extra directory existence check which should filter out non-existent or empty directories.

[kbrock]

ooh, sweet!

Ugh, I guess I have been living in the homebrew world and forgot about version 1.0.

[postmodern]

I've been slowly working on 1.0.0 branch where I can make breaking changes. The last feature I was trying to work on is adding a auto.sh like feature but for setting up aliases for bundle exec commands; had trouble finding all executables from bundler's API. I also plan on eventually releasing a 0.4.0 which adds some additional features and relieves some of the pain some users are feeling, such as the slow load times.

[kbrock]

1. Just to be clear, 3.9 is plenty fast. This all started for me because nvm is slow and I was looking at chruby as an example of how to make nvm as quick as chruby.
2. Most people I know have a be or ber/beer (bundle exec and bundle exec ruby). I went a little too far and have a "smart" beer command that calls ruby, rails, rake, or rspec depending upon the parameters. Since rails introduce something similar to their rails command, it seems they felt a similar pain. Wonder if the added lines of code justify the additional functionality in chruby?