Recently one of the remotely maintained servers I worked on regularly has started to exhibit slowdowns in bash. It started only occasionally, perhaps as a result of a high number of users or long-running script, but become a lot worse and resulted in a few seconds delay every time a new prompt was drawn. A few seconds may not sound a lot, but can be incredibly frustrating!
Upon a bit of an investigation I found that the culprit was the call to
__git_ps1 in my
$PS1 variable. Removing this made the bash prompt speedy again! Hooray! However, having the current branch on my screen immediately is infinitely useful and I can’t really live without it any more, so I carried on digging.
On this particular set of servers
git has been replaced with a set of custom script that perform a lot of tasks. This means that calls to simple
git commands run a LOT of other commands, logging and getting repo information among other things. I thought about replacing all the references to
git in the
git-prompt.sh file with the direct path to the binary, but thought it shouldn’t be too tough to just get the information myself using the information in the
The result is this bash function.
It’s fairly primitive, it checks the current path for a
.git directory, if it doesn’t find one, it goes up one level, until either finding one, or hitting the root. It
printfs the result in the same way as the real
__git_ps1 and handles the basic states like
|MERGING although doesn’t detect a detached HEAD any other similar states, but serve the main purpose, now I can correctly see my current branch and avoid committing the wrong thing to the wrong place!
Feel free to share if this is useful and let me know if there are any minor improvements (short of using the full version!).
greping over a codebase, there can be so many different things to trip you up and flag false positives, or files that you definitely didn’t want to
Continue reading “
grep: defaults for web developers”
I’ve always noticed really elaborate examples of what can be achieved with a highly customised
PS1 variable for your bash prompt and recently have spent some time playing with my settings to make my terminal experience better with a few updates. I also felt that I would have benefitted in the past from being able to analyse what my prompt currently does, breaking it down into easier to digest blocks, so I can make more sense of what each part does.
As a result of that curiosity, I’ve created a small page that allows you to paste in an existing
PS1 from your terminal which will then be parsed and previewed with some basic dummy details (which can be changed by clicking on the identifier in the preview area).
I’d like to make it into something like EzPrompt with support for 256, or even true colour additions and collate some useful scripts for displaying additional information, but at the moment it’s useful for seeing what each block does at least.
It seems to cover most of the examples I’ve tested in various searches, but the more inline-code is included, the less it can display.
Check it out here and the source is available on github.
I’ve been working on cleaning up my dotfiles recently and have been playing with my
PS1 and the available colours. In doing so I thought it might be interesting to extract image data from an uploaded image and turn the result into a nearest-match
ANSI image. This is the result! It supports both 256 colour and true colour terminals and can utilise unicode code points to present a clear image in slightly fewer pixels as well as just using background colours and spaces if preferred.
It includes a live preview which I’ve lifted from the bash
Check it out here!