* refactor(bash): use `STARSHIP_*` for the internal variable names
The current codebase uses `dbg_trap` to save the original DEBUG trap
in bash <= 4.3. However, the variable name possibly conflicts a user
variable since it is not prefixed by `_starship` or `starship_` or
`STARSHIP_`. In this patch, we rename `dbg_trap` to
`STARSHIP_DEBUG_TRAP` following other variables of
`STARSHIP_EXIT_STATUS` and `STARSHIP_PIPE_STATUS`. We also rename the
variable `_PRESERVED_PROMPT_COMMAND` to `STARSHIP_PROMPT_COMMAND`.
* fix(bash): correctly extract DEBUG trap with spaces
* fix(bash): evaluate DEBUG trap by eval
The current code executes the saved DEBUG trap just by
$_starship_dbg_trap. This causes various problems. The content of
the variable `_starship_dbg_trap` is unexpectedly subject to the word
splitting with the pathname expansions. Also, the needed shell
expansions and quote removal are not performed. With a custom IFS,
the content of the variable will be split in an unexpected way. The
saved DEBUG trap needs to be executed by eval "$_starship_dbg_trap".
* feat(k8s): Add detect env vars option
Have added the option to trigger the k8s module based on what env vars
are set, this has been done in a backwards compatible way so if nothing
is changed from the defaults the module will still behave the same way
as before. This is similar to what I did in #4486 for the python module
and if goes well I'd like to rollout to other modules.
* Update src/modules/kubernetes.rs
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
* Update src/modules/kubernetes.rs
---------
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
* combine ANSI color codes before wrapping them
The existing code wraps each individual module's output for
`context.shell`, concatenates all that output together and passes it to
`AnsiStrings` to merge ANSI color codes. However, the wrapping obscures
ANSI color codes, meaning that no merging is possible.
This commit changes the shell-specific wrapping to happen right before
output, once all modules' output has been concatenated together. This
results in ANSI color codes being correctly merged, as well as reducing
the number of calls to `wrap_colorseq_for_shell` to one.
With a minimal `starship.toml`:
```
format = """$directory"""
[directory]
format = '[a]($style)[b]($style)'
```
The current code produces[0]:
```
\n%{\x1b[31m%}a%{\x1b[0m%}%{\x1b[31m%}b%{\x1b[0m%
```
And this commit's code:
```
\n%{\x1b[31m%}ab%{\x1b[0m%}
```
You can see that the current code emits an additional reset and repeated
color code between "a" and "b" compared to the new code.
[0] Produced in a Python shell with:
```
subprocess.check_output(
"./target/debug/starship prompt", shell=True,
env={"STARSHIP_CONFIG": "./starship.toml", "STARSHIP_SHELL": "zsh"}
)
```
* utils: return early from wrap_seq_for_shell unless wrapping required
* refactor(utils): simplify wrap_seq_for_shell
This commit modifies wrap_seq_for_shell to (a) return early for shells
with no wrapping required, and (b) determine the wrapping characters
once at the start of the function (rather than inline in the map
function for every character).
