Not sure I’d call what bash has functions. They’re closer to subroutines in Basic than functions in other languages, as in you can’t return a value from them (they can only return their exit code, and you can capture their stdout and stderr). But even then, they are full subshells. It’s one of the reasons I don’t really like Bash, you’re forced into globally or at least broadly-scoped variables. Oh, and I have no clue right now how to find where in your pipe you got a non-null exit code.
It’s not a big problem for simple scripting, but it makes things cumbersome once you try to do more.
You’re not forced into global forced variables, but they’re the default. Use the local keyword in front of the variable declaration for nicely scoped variable.
It’s not that cumbersome to do things like
localdate=`date`
echo"$date"
but in all honesty the syntax sucks ass because it’s not intuitive. If statements suck ass, passing variables has to be done via command line arguments sucks ass, switch statements suck ass, making structured data sucks ass (jq is nice though).
I agree with you that bash really sucks when you get to anything more than 10 lines and at that point I’d take literally prefer Dreamberd.
First thing you want is set -e and set -o pipefail. That should report the errors in human-parseable form.
Second, to capture exit codes from each command/program, you have to run each of them in sequence yourself, connected by pipes that you create via mkfifo — the same way as you would do it in any other programming environment. Bash’s | pipes are just a convenient shorthand for this,so if you want full control, you have to ditch the convenience.
Functions are definitely not subshells in Bash, seeing as anything modifying the environment, like pyenv and such, is implemented as functions instead of scripts — specifically because functions are run in the same shell instance.
Unless ‘subshell’ means something in the vein of ‘like a new shell, but not really’.
I’m gonna bet yes for the simple reason that various helper scripts exist that do advanced cd history, with fuzzy search and whatnot, and they can’t be implemented as anything other than functions.
I really like bash when dealing with even somewhat advanced scripting. Like the 300 LOC scraper I have written over the past two days which horribly parses HTML files using grep | sed.
It’s genuinely so much more fun to do this with Bash than, say, Python. I have once written a scraper using Beautifulsoup and I have no desire to do so ever again.
Honestly, only Haskell manages to beat Bash in how satisfying it feels when you manage to get something working well.
Bash has its upsides too, like the fact that it has arrays / lists and dictionaries / hashmaps. In my opinion, it gets iffy though when you need to do stuff with IFS; at that point one might be better off just using specialized tools.
Not saying working bash isn’t good enough, but it can break in very surprising ways is my experience.
Not sure I’d call what bash has functions. They’re closer to subroutines in Basic than functions in other languages, as in you can’t return a value from them (they can only return their exit code, and you can capture their stdout and stderr). But even then, they are full subshells. It’s one of the reasons I don’t really like Bash, you’re forced into globally or at least broadly-scoped variables. Oh, and I have no clue right now how to find where in your pipe you got a non-null exit code.
It’s not a big problem for simple scripting, but it makes things cumbersome once you try to do more.
You’re not forced into global forced variables, but they’re the default. Use the
localkeyword in front of the variable declaration for nicely scoped variable.It’s not that cumbersome to do things like
local date=`date` echo "$date"but in all honesty the syntax sucks ass because it’s not intuitive. If statements suck ass, passing variables has to be done via command line arguments sucks ass, switch statements suck ass, making structured data sucks ass (
jqis nice though).I agree with you that bash really sucks when you get to anything more than 10 lines and at that point I’d take literally prefer Dreamberd.
First thing you want is
set -eandset -o pipefail. That should report the errors in human-parseable form.Second, to capture exit codes from each command/program, you have to run each of them in sequence yourself, connected by pipes that you create via
mkfifo— the same way as you would do it in any other programming environment. Bash’s|pipes are just a convenient shorthand for this,so if you want full control, you have to ditch the convenience.Functions are definitely not subshells in Bash, seeing as anything modifying the environment, like
pyenvand such, is implemented as functions instead of scripts — specifically because functions are run in the same shell instance.Unless ‘subshell’ means something in the vein of ‘like a new shell, but not really’.
You’re right, my bad, I got this mixed up with something else.
Try doing
cdinside a function and thenpwdin the main script. Does it move the main script too?I’m gonna bet yes for the simple reason that various helper scripts exist that do advanced
cdhistory, with fuzzy search and whatnot, and they can’t be implemented as anything other than functions.I really like bash when dealing with even somewhat advanced scripting. Like the 300 LOC scraper I have written over the past two days which horribly parses HTML files using grep | sed.
It’s genuinely so much more fun to do this with Bash than, say, Python. I have once written a scraper using Beautifulsoup and I have no desire to do so ever again.
Honestly, only Haskell manages to beat Bash in how satisfying it feels when you manage to get something working well.
Bash has its upsides too, like the fact that it has arrays / lists and dictionaries / hashmaps. In my opinion, it gets iffy though when you need to do stuff with IFS; at that point one might be better off just using specialized tools.
Not saying working bash isn’t good enough, but it can break in very surprising ways is my experience.