Expand description
§Overview
This is a rust attribute-like proc macro which reduces the amount of code required to call shell commands and parse the results.
It allows you to wrap a script in any language with strongly typed functions. The function’s arguments are set as env variables and the result of the script is parsed either as a value or as an iterator.
§Examples
§Basic
use shellfn::shell;
use std::error::Error;
#[shell]
fn list_modified(dir: &str) -> Result<impl Iterator<Item=String>, Box<Error>> { r#"
cd $DIR
git status | grep '^\s*modified:' | awk '{print $2}'
"# }§Different interpreter
use shellfn::shell;
use std::error::Error;
#[shell(cmd = "python -c")]
fn pretty_json(json: &str, indent: u8, sort_keys: bool) -> Result<String, Box<Error>> { r#"
import os, json
input = os.environ['JSON']
indent = int(os.environ['INDENT'])
sort_keys = os.environ['SORT_KEYS'] == 'true'
obj = json.loads(input)
print(json.dumps(obj, indent=indent, sort_keys=sort_keys))
"# }§Usage
You can use the #[shell] attribute on functions that have:
- a body containing only one expression - a string literal representing the script to execute
- types that implement the
.to_string()method - return a value that is either
void,T,Result<T, E>,impl Iterator<Item=T>,Result<impl Iterator<Item=T>>orResult<impl Iterator<Item=Result<T, E>>>with constrains:
T: FromStr,
<T as FromStr>::Err: StdError,
E: From<shellfn::Error<<T as FromStr>::Err>>,§Details
The #[shell] attribute does the following:
- Sets every argument as an env variable
- Runs a shell command
- Launches the command using
std::process::Command - Depending on the return type, it may parse the output
Most of the steps can be adjusted:
- the default command is
bash -c. You can change it using thecmdparameter:
#[shell(cmd = "python -c")]- by default, the script is added as the last argument. You can change it using the special variable
PROGRAMin thecmdparameter:
#[shell(cmd = "bash -c PROGRAM -i")]- you can use env variables set from function’s arguments in the
cmdparameters in the same way as in the script:
#[shell(cmd = "python -m $MODULE")]
fn run(module: &str)- if the return type is not wrapping some part of the result in
Result, you may decide to suppress panics by adding theno_panicflag:
#[shell(no_panic)]Following return types are currently recognized:
| return type | flags | on parse fail | on error exit code | on spawn fail | notes |
|---|---|---|---|---|---|
| - | panic | panic | |||
| no_panic | - | nothing | nothing | ||
| () | - | panic | panic | ||
| () | no_panic | - | nothing | nothing | |
| Result<(), E> | - | error | error | ||
| Result<(), E> | no_panic | - | error | error | 1 |
| T | panic | panic | panic | 2 | |
| T | no_panic | panic | panic | panic | 1,2 |
| Result<T, E> | error | error | error | 2 | |
| Result<T, E> | no_panic | error | error | error | 1,2 |
| Vec | panic | panic | panic | ||
| Vec | no_panic | skip | ignored | empty vec | 3 |
| Vec<Result<T, E>> | item error | panic | panic | ||
| Vec<Result<T, E>> | no_panic | item error | ignored | empty vec | |
| Result<Vec | panic | error | error | ||
| Result<Vec | no_panic | skip | error | error | |
| Result<Vec<Result<T, E1>>, E2> | item error | error | error | ||
| Result<Vec<Result<T, E1>>, E2> | no_panic | item error | error | error | 1 |
| impl Iterator<Item=T> | panic | panic | panic | ||
| impl Iterator<Item=T> | no_panic | skip | ignored | empty iter | 3 |
| impl Iterator<Item=Result<T, E>> | item error | panic | panic | 3 | |
| impl Iterator<Item=Result<T, E>> | no_panic | item error | ignored | empty iter | |
| Result<impl Iterator<Item=T>, E> | panic | ignored | error | ||
| Result<impl Iterator<Item=T>, E> | no_panic | skip | ignored | error | |
| Result<impl Iterator<Item=Result<T, E1>>, E2> | item error | ignored | error | ||
| Result<impl Iterator<Item=Result<T, E1>>, E2> | no_panic | item error | ignored | error | 1 |
Glossary:
| action | meaning |
|---|---|
| panic | panics (.expect or panic!) |
| nothing | consumes and ignores error (let _ = …) |
| error | returns error |
| skip | yields all successfuly parsed items, ignores parsing failures (filter_map) |
| empty iter/vec | returns empty iterator / vector |
| item error | when parsing fails, yields Err |
| ignored | ignores exit code, behaves in the same way for exit code 0 and != 0 |
Notes:
- The
no_panicattribute makes no difference - It reads all of stdout before producing any failures
- It yields all items until it encounters an error or an exit code
§Vector vs iterator
Variants with the Vec return type are very similar to the ones with impl Iterator. The key differences are:
impl Iteratoris only allocating one item at the time and yields it immediately after it is parsed, whileVecis reading output line by line but stores parsed output in the temporary VecVecis aware of exit code. When subprocess finishes with error,impl Iteratorwill stop yielding values whileVecwill return error or panic
Enums§
Functions§
- execute_
iter_ nopanic_ nopanic - Executes command with args and environment variables, parses output line by line
- execute_
iter_ nopanic_ result - Executes command with args and environment variables, parses output line by line
- execute_
iter_ panic_ panic - Executes command with args and environment variables, parses output line by line
- execute_
iter_ panic_ result - Executes command with args and environment variables, parses output line by line
- execute_
iter_ result_ nopanic - Executes command with args and environment variables, parses output line by line
- execute_
iter_ result_ panic - Executes command with args and environment variables, parses output line by line
- execute_
iter_ result_ result - Executes command with args and environment variables, parses output line by line
- execute_
parse_ panic - Executes command with args and environment variables, parses output
- execute_
parse_ result - Executes command with args and environment variables, parses output
- execute_
vec_ nopanic_ nopanic - Executes command with args and environment variables, parses output line by line, returns after reading whole output
- execute_
vec_ nopanic_ result - Executes command with args and environment variables, parses output line by line, returns after reading whole output
- execute_
vec_ panic_ panic - Executes command with args and environment variables, parses output line by line, returns after reading whole output
- execute_
vec_ panic_ result - Executes command with args and environment variables, parses output line by line, returns after reading whole output
- execute_
vec_ result_ nopanic - Executes command with args and environment variables, parses output line by line, returns after reading whole output
- execute_
vec_ result_ panic - Executes command with args and environment variables, parses output line by line, returns after reading whole output
- execute_
vec_ result_ result - Executes command with args and environment variables, parses output line by line, returns after reading whole output
- execute_
void_ nopanic - Executes command with args and environment variables, ignores output
- execute_
void_ panic - Executes command with args and environment variables, ignores output
- execute_
void_ result - Executes command with args and environment variables, ignores output