* refactor: replace module_config_derive with serde
Changes include:
* Removing `starship_module_config_derive` and replacing it with `serde::Deserialize`
* Removing `RootModuleConfig::load_config`. While potentially useful, it was only used in tests. And it would require something like `serde::DeserializeSeed` which is not derived by serde.
* Merging `RootModuleConfig` into `ModuleConfig`
* Implementing a `ValueDeserializer` that holds a reference to a `toml::Value` in `serde_utils.rs`
* Deserialization errors (invalid type) are now logged and include the current key and the struct names
* Unknown keys are now considered an error. "Did you mean?"-messages are still possible
* fix typo
Co-authored-by: Matan Kushner <hello@matchai.dev>
Co-authored-by: Matan Kushner <hello@matchai.dev>
* add option to force AWS display
Even if no credentials or credential_process have been setup
* change README wording
* Include sso_start_url in the description
* Change option name to force_display
* feat(haskell): add haskell module and implement ghc version detection
* feat(haskell): implement stack resolver version detection
* fix(haskell): handle more complex resolvers
* fix(haskell): rename resolver_version to snapshot
* fix(haskell): change default color to bold purple
* feat(haskell): add tests
* fix(haskell): format
* fix(haskell): replace incorrect `or` with `or_else`
* fix(haskell): use write_all instead of write
* fix(haskell): λ as Haskell icon by default
* fix(haskell): fix tests and add a real stack.yaml testcase
* fix(haskell): make clippy happy
* Rename m.aws.alias_region to alias_name
* Add aws profile aliases
* Document aws.profile_aliases, with examples
* Add tests for new aws.profile_aliases feature
* Tidy alias_handling a bit
Fixes a bug in the Rust module where overrides would match
against both full path components as well as partial ones
(e.g. "/etc/passwd" would be considered a prefix of
"/etc/passwd_new") on Windows.
Solution is to convert back to a PathBuf after lossy conversion
so that Path::starts_with is used instead of str::starts_with.
* only display aws on credential_process defined
* add check for both credential_process and valid credentials
* fix tests
* update aws module documentation
* add better explanation of requirements to documentation
* add support for AWS_SECRET_ACCESS_KEY, AWS_SESSION_TOKEN, and update docs
* remove credential_process env var
* Use global:error[0] for most recent error
The current version use $error[0], which is for the module instead of the global status and is always $null, so the $lastExitCodeForPrompot is always $origLastExitCode (see https://github.com/starship/starship/issues/3485 for a related bug).
It should be replaced with $global:error[0]
Comments are also updated
* fix(pwsh): make Semantic PR bot happy
* fix(#3554): Print the command line argv on clap error
This is a very bare implementation that just prints the error
and then a note with the arguments passed, it does this manually
and doesn't use clap. I've also chosen to use `Vec`'s `Debug`
implementation instead of rolling my own one because I thought it was
good enough, but there might be a better way of doing all this.
Altogether, I think this will be very useful to help in the diagnostic
of other bugs :)
* fix(#3554): Print the command line argv on clap error
This is a very bare implementation that just prints the error
and then a note with the arguments passed, it does this manually
and doesn't use clap. I've also chosen to use `Vec`'s `Debug`
implementation instead of rolling my own one because I thought it was
good enough, but there might be a better way of doing all this.
Altogether, I think this will be very useful to help in the diagnostic
of other bugs :)
EDIT: removed `dbg!`, set it to exit always.
* correctness(exit): don't print argv / exit with error on help and
version error kinds
* fix: Avoid panicking when stdout/stderr closing unexpectedly
* refactor(cli): use `use_stderr` instead of manual match for error kinds
`clap` uses `use_stderr` to reliably check whether the error given is
actually an error coming from user input or rather a hint to display
other info (version, help, etc.)
Also reworded/moved a couple of comments so that they explain better
what is the thought process behind the code
* fix(status): Enable to convert from i64 to hex_status by casting instead of parsing status.
* Apply comment to src/context.rs
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
* Update README.md in configuration
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
In #1019 the undistract feature has been added to starship without
enabling it by default due to the system dependency on libdbus. With
recent versions of notify-rust the dependency is no longer required and
the feature can be enabled by default.
* fix: use e718 as the default of symbol in node configuration
* wip: change nodejs symbol in docs/config & add a nodejs symbol configuration in docs/presets
* wip: update CONTRIBUTING.md
* Allow customization of notification timeout
* Document new notification duration option
* Check for out-of-bounds timeout and correct it
* Implement ModuleConfig for u32
* Revert "Check for out-of-bounds timeout and correct it"
This reverts commit 52109ab5f7c336b55c81bccafb3adbfc81514553.
* Switch notification_timeout to u32
* Note notification_daemons might not honor timout
* Notification timeout defaults to daemon timeout
* Leave default value of notification_timeout blank in docs
The `$PWD` environment variable used by starship is not updated on startup by `elvish`. `elvish` also provides a separate `$pwd` variable that does get updated and seems to be be a better indicator for the current logical path.
* ci: Fix aws::expiration_date_set_from_file race
While aws::expiration_date_set_from_file will almost-always work
perfectly locally, it is theoretically possible for the thread running
the test to be scheduled away betwen writing the file with timing
information and then actually reading it, resulting in a
shorter-than-expected time appearing in the module. This can also happen
if the test triggers right at the very end of a second (e.g. at
10:45:47.999995).
This appears to actually happen sometimes on heavily-loaded GitHub
Actions runners. To fix this issue, we allow for up to a two-second
delay between when the file is written and when the test actually fires
(allowing "30m", "29m59s", and "29m58s" as possible values).
* Fix typo
* test: add mock method for absolute files
Signed-off-by: Harald Hoyer <harald@hoyer.xyz>
* feat(module): add a container indicator module
Adds a container type indicator, if inside a container,
detected via the presence of some marker files.
E.g. inside a podman container entered with `toolbox enter`
the prompt changes to the container name and version.
```
starship on container_rebased [$!] is 📦 v1.0.0 via 🦀 v1.56.1
❯ toolbox enter
starship on container_rebased [$!] is 📦 v1.0.0 via 🦀 v1.56.1
⬢ [fedora-toolbox:35] ❯
```
Signed-off-by: Harald Hoyer <harald@hoyer.xyz>
* feat(pwsh): Set STARSHIP_SHELL to pwsh ON PS >5
This is needed for the custom module to use the correct shell.
Closes #3376
* Update src/context.rs
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
* fix(pwsh): Avoid polluting the global function namespace
This is an alternative to #3386
* Review changes
* Move continuation prompt after setting starship env
* feat: add support for cmd
* add preprompt and precmd support
* add keymap support
* add info about minimum Clink version
* simplify escaping
* add handling for cmd custom commands
* add support for transient_prompt and transient_rprompt
* Revert 914057952508e81e20086fcb707ba2a0be85fdd3
This reverts commit "add support for transient_prompt and transient_rprompt"
* Apply suggestions from code review
* disable cmd shell custom commands
* any shell other than cmd can be used
* better error and correct script location
* move shell check in `map_no_escaping`
* feat: set a continuation prompt for supporting shells (#3134)
* docs: fixed wording of documentation
* fix: continuation prompt is now only set once
* fix(docs): fixed typo in advanced-config/README.md
Co-authored-by: Segev Finer <segev208@gmail.com>
* fix: update --continuation argument
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
* fix: updated continuation prompt
- PROMPT2 was fixed to be set only once in zsh.
- `continuation_symbol` and `continuation_format` were removed in
place of a single variable; `continuation_prompt`.
- The continuation prompt was moved out of the character module.
* fix: ran rustfmt
* docs: updated continuation prompt docs
Co-authored-by: Segev Finer <segev208@gmail.com>
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
* perf(git_status): tweak flags to omit extra info
`git status` can be prohibitively slow on some repos, so allow the
config to influence what flags are passed to git. For instance, if there
is no configured symbol for untracked files, tell git to omit them from
its output. This can easily result in a 2~10x speedup in many cases, but
requires the user to opt-in to hiding information from the prompt.
* docs(git_status): add ignore_submodules option
* feat: Experimental Windows path formatting via path_slash::PathBufExt
* Rework the slash path conversion into a real PR
* Add a test for convert_slash = false
* Attempt fixing CI failures
* Fix lint and fmt
* Fix docs/config/README.md getting messed up
* Rename convert_slash/from_slash
* Move convert_path_sep calls in tests
* Keep path_vec immutable
* Run rustfmt
* perf(rust): additionally check `rustup default` for faster result.
After checking directory overrides we were directly falling back to the
relatively slow call to `rustc --version`.
Inserting a call to `rustup default` leads to a quicker response.
* use `context.exec_cmd` instead of `create_command`
Previous code would render all components as empty, resulting
in an empty string even if min_time was 0. This adds a special
case which forces prompt to render "0ms"
When opening a directory as a file the intial open works, while
subsequent line-reads will fail with _is a directory_.
Since erroring line-reads were just skipped this lead to an
endless loop.
These need to be stored before calling any PowerShell function or executable.
Otherwise the values will be overwritten and cannot be properly restored
or passed to starship.
Fixes: #3315
* add feature - sudo module
* add sudo module identifiers and entry point
* fix test test_sudo_not_cached
* add test test_sudo_cached
* add `allow_windows` and `binary` options
* rustfmt sudo_x_cached and rmv them on windows
* add false `allow_windows` block windows test
* add `doas` cached/not_cached tests
* better description in `starship explain`
* fix `test_doas_cached` with `-n` flag
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
* rmv `binary` alternatives and their tests
* fix symbol and update config/README
* fix all mocks to use `sudo -n true`
* fix expected output in `test_sudo_cached`
* proper checking for blocked sudo
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
* add `allow_windows = true` to non-windows tests
* allow sudo_* tests to run on windows + fix parsed
* rustfmt `blocks_windows` test
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
This should fix some modules not working correctly in the fish
`RPROMPT`. I have done this the very naive way by simply duplicating the
code that is currently in the `fish_prompt` function to avoid having to
change the scope of any of the variables (currently all local).
* add crystal shard (package) version support
* module package: crystal shard version: read shard.yml directly
* module package: add test for crystal shard version
* format src/modules/package.rs
* use yaml-rust instead of serde-yaml
* document shards package support
* Update docs/config/README.md
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
Co-authored-by: David Knaack <davidkna@users.noreply.github.com>
* allow unset KEYMAP and STARSHIP_DURATION in zsh
which prevents errors if a user has `set -u` on in there terminal.
* fix: allow unset KEYMAP and STARSHIP_DURATION in zsh
Add `terminal-width` argument to explicitly define terminal's width
Update pwsh,bash,zsh,fish init scripts with `terminal-width` argument
Co-authored-by: Kevin Song <chips@ksong.dev>
Co-authored-by: Kevin Song <4605384+chipbuster@users.noreply.github.com>
Changes the parsing for pipestatus to allow for multiple arguments, a
single argument of space-separated values, or any mix of the two. All
inputs are flattened into a single array where no elements have spaces
in them.
Changes the initscripts to no longer fail when an empty pipestatus
is passed as an argument by quoting and changing expansions.
Have updated which extentions trigger the terraform module, to remove
the `hcl` since this gives false positives as it is used by other
Hashicorp products and add the `tfstate` and `tfplan` which are more
accurate indicators.
Due to the introduction of utils::create_command, commands now have
stdin set to null, and stdout and stderr set to piped.
This prevents console editors from working when invoked via
starship config
Have added version formatting to the red and vlang modules. Note the
docs for red already mentioned the `version_format` string but it had
not actually been added.
* feat: Add the symbol and number thresholds respecting the threshold option
* fix: Maintain the old behavior + add lots of tests
* docs: Fix the jobs module documentation
* git_status: added symbol for when local branch is up-to-date with upstream
* updated docs
* removed unused variable, moved location of config comment
* changed uptodate default to empty string, simplified and made safer
* added uptodate default line back into docstring
* fixed linting and formatting errors
* refactored uptodate to up_to_date, removed redundant else statement
Have added the ability to use format the version of the package using
the `version_format` option. While doing this I have also done some
refactoring of the module to remove the if/else if/... block and replace
it with an iterator. This should make fix some edge cases where versions
are not correctly picked up due to other files an example would be a
python project that has a `pyproject.toml` file but using the
`setup.cfg` for the package version. It should also make it easier to
make the order of the list configurable in the future.
* feat: Add pipestatus display in status module
This MR is based on this one https://github.com/starship/starship/pull/370
* Documentation
* Add a test with map_symbol false
* Handle bash preexec pipestatus
* Add zsh support
* Add fish support
Thanks kidonng for the diff patch
* Rename sucess_symbol to success_symbol
* Support package version from setup.cfg (python).
Add an additional package version extraction function to parse
the 'version' attribute under the 'metadata' section in a python
package 'setup.cfg' file.
Also add similar tests from the poetry extraction function to test
the desired behaviour.
This adds a dependency on ConfigParser:
https://crates.io/crates/configparser.
* Clean up comments
* Use rust_ini over ConfigParser
* Add mention to setup.cfg version parsing in docs
* feat: add support for xonsh
* xonsh: add STARSHIP_SESSION_KEY
* xonsh: implement STARSHIP_SESSION_KEY in xonsh
* docs: mention tcsh, elvish, and nu in more places
* xonsh: change STARSHIP_SESSION_KEY implementation
See https://github.com/starship/starship/pull/2807#discussion_r667064149
* xonsh: fix jobs implementation
* xonsh: do not silently discard stderr from starship
On Windows when running commands with their name instead of the path with Command::new, executable with that name from the current working directory will be executed.
This PR replaces all instances of Command::new with a new create_command function which will first resolve any executable paths and avoid this issue.
