Turn Clojure functions into CLIs!
This library is still in design phase and may still undergo breaking changes. Check breaking changes before upgrading!
Add to your deps.edn
or bb.edn
:deps
entry:
org.babashka/cli {:mvn/version "0.3.35"}
Command line arguments in clojure and babashka CLIs are often in the form:
$ cli command :opt1 v1 :opt2 v2
or the more Unixy:
$ cli command --long-opt1 v1 -o v2
The main ideas:
- Put as little effort as possible into turning a clojure function into a CLI,
similar to
-X
style invocations. For lazy people like me! If you are not familiar withclj -X
, read the docs here. - But with a better UX by not having to use quotes on the command line as a
result of having to pass EDN directly:
:dir foo
instead of:dir '"foo"'
or who knows how to write the latter incmd.exe
or Powershell. - Open world assumption: passing extra arguments does not break and arguments can be re-used in multiple contexts.
- Because the line between calling functions from the command line and Clojure itself is blurred, validation of arguments should happen in your Clojure function, using your favorite tools (manually, spec, schema, malli...). As such, the library only focuses on coercion: turning argument strings into data which is then passed to your function.
Both :
and --
are supported as the initial characters of a named option, but
cannot be mixed. See options for more
details.
See clojure CLI for how to turn your exec functions into CLIs.
For parsing options, use either parse-opts
or parse-args
.
Examples:
Parse {:port 1339}
from command line arguments:
(require '[babashka.cli :as cli])
(cli/parse-opts ["--port" "1339"] {:coerce {:port :long}})
;;=> {:port 1339}
Use an alias (short option):
(cli/parse-opts ["-p" "1339"] {:alias {:p :port} :coerce {:port :long}})
;; {:port 1339}
Coerce values into a collection:
(cli/parse-opts ["--paths" "src" "--paths" "test"] {:coerce {:paths []}})
;;=> {:paths ["src" "test"]}
(cli/parse-opts ["--paths" "src" "test"] {:coerce {:paths []}})
;;=> {:paths ["src" "test"]}
Transforming to a collection of a certain type:
(cli/parse-opts ["--foo" "bar" "--foo" "baz"] {:coerce {:foo [:keyword]}})
;; => {:foo [:bar :baz]}
Booleans need no explicit true
value and :coerce
option:
(cli/parse-opts ["--verbose"])
;;=> {:verbose true}
(cli/parse-opts ["-v" "-v" "-v"] {:alias {:v :verbose}
:coerce {:verbose []}})
;;=> {:verbose [true true true]}
Long options also support the syntax --foo=bar
:
(cli/parse-opts ["--foo=bar"])
;;=> {:foo "bar"}
Since v0.3.35
babashka CLI auto-coerces values that have no explicit coercion
with
auto-coerce
:
it automatically tries to convert booleans, numbers and keywords.
To parse positional arguments, you can use parse-args
and/or the :args->opts
option. E.g. to parse arguments for the git push
command:
(cli/parse-args ["--force" "ssh://foo"] {:coerce {:force :boolean}})
;;=> {:args ["ssh://foo"], :opts {:force true}}
(cli/parse-args ["ssh://foo" "--force"] {:coerce {:force :boolean}})
;;=> {:args ["ssh://foo"], :opts {:force true}}
Note that this library can only disambiguate correctly between values for
options and trailing arguments with enough :coerce
information
available. Without the :force :boolean
info, we get:
(cli/parse-args ["--force" "ssh://foo"])
{:opts {:force "ssh://foo"}}
In case of ambiguity --
may also be used to communicate the boundary between
options and arguments:
(cli/parse-args ["--paths" "src" "test" "--" "ssh://foo"] {:coerce {:paths []}})
{:args ["ssh://foo"], :opts {:paths ["src" "test"]}}
To fold positional arguments into the parsed options, you can use :args->opts
:
(def cli-opts {:coerce {:force :boolean} :args->opts [:url]})
(cli/parse-opts ["--force" "ssh://foo"] cli-opts)
;;=> {:force true, :url "ssh://foo"}
(cli/parse-opts ["ssh://foo" "--force"] cli-opts)
;;=> {:url "ssh://foo", :force true}
If you want to fold a variable amount of arguments, you can coerce into a vector
and specify the variable number of arguments with repeat
:
(def cli-opts {:coerce {:bar []} :args->opts (cons :foo (repeat :bar))})
(cli/parse-opts ["arg1" "arg2" "arg3" "arg4"] cli-opts)
;;=> {:foo "arg1", :bar ["arg2" "arg3" "arg4"]}
Use the :restrict
option to restrict options to only those explicitly mentioned in configuration:
(cli/parse-args ["--foo"] {:restrict [:bar]})
;;=>
Execution error (ExceptionInfo) at babashka.cli/parse-opts (cli.cljc:357).
Unknown option: :foo
Use the :require
option to throw an error when an option is not present:
(cli/parse-args ["--foo"] {:require [:bar]})
;;=>
Execution error (ExceptionInfo) at babashka.cli/parse-opts (cli.cljc:363).
Required option: :bar
(cli/parse-args ["--foo" "0"] {:validate {:foo pos?}})
Execution error (ExceptionInfo) at babashka.cli/parse-opts (cli.cljc:378).
Invalid value for option :foo: 0
To gain more control over the error message, use :pred
and :ex-msg
:
(cli/parse-args ["--foo" "0"] {:validate {:foo {:pred pos? :ex-msg (fn [m] (str "Not a positive number: " (:value m)))}}})
;;=>
Execution error (ExceptionInfo) at babashka.cli/parse-opts (cli.cljc:378).
Not a positive number: 0
You can supply default args with :exec-args
:
(cli/parse-args ["--foo" "0"] {:exec-args {:bar 1}})
;;=> {:foo 0, :bar 1}
Note that args specified in args
will override defaults in :exec-args
:
(cli/parse-args ["--foo" "0" "--bar" "42"] {:exec-args {:bar 1}})
;;=> {:foo 0, :bar 42}
By default, an exception will be thrown in the following situations:
- A restricted option is encountered
- A required option is missing
- Validation fails for an option
- Coercion fails for an option
You may supply a custom error handler function with :error-fn
. The function
will be called with a map containing the following keys:
:type
-:org.babashka/cli
(for filtering out other types of errors).:cause
- one of::restrict
- a restricted option was encountered.:require
- a required option was missing.:validate
- validation failed for an option.:coerce
- coercion failed for an option.
:msg
- default error message.:option
- the option being parsed when the error occurred.:spec
- the spec passed intoparse-opts
(see the Spec section).
The following keys are present depending on :cause
:
:cause :restrict
:restrict
- the value of the:restrict
opt toparse-args
(see the Restrict section).
:cause :require
:require
- the value of the:require
opt toparse-args
(see the Require section).
:cause :validate
:value
- the value of the option that failed validation.:validate
- the value of the:validate
opt toparse-args
(see the Validate section).
:cause :coerce
:value
- the value of the option that failed coercion.
It is recommended to either throw an exception or otherwise exit in the error
handler function, unless you want to collect all of the errors and act on them
in the end (see babashka.cli-test/error-fn-test
for an example of this).
For example:
(cli/parse-opts
[]
{:spec {:foo {:desc "You know what this is."
:ref "<val>"}}
:error-fn
(fn [{:keys [spec type cause msg option] :as data}]
(if (= :org.babashka/cli type)
(case cause
:require
(println
(format "Missing required argument:\n%s"
(cli/format-opts {:spec (select-keys spec [option])})))
(println msg))
(throw (ex-info msg data)))
(System/exit 1))})
would print:
Missing required argument:
--foo <val> You know what this is.
This library can work with partial information to parse options. As such, the
options to parse-opts
and parse-args
are optimized for terseness. However,
when writing a CLI that supports automated printing of options, it is recommended to use the spec format:
(def spec {:from {:ref "<format>"
:desc "The input format. <format> can be edn, json or transit."
:coerce :keyword
:alias :i
:default-desc "edn"
:default :edn}
:to {:ref "<format>"
:desc "The output format. <format> can be edn, json or transit."
:coerce :keyword
:alias :o
:default-desc "json"
:default :json}
:pretty {:desc "Pretty-print output."
:alias :p}
:paths {:desc "Paths of files to transform."
:coerce []
:default ["src" "test"]
:default-desc "src test"}})
You can pass the spec to parse-opts
under the :spec
key: (parse-opts args {:spec spec})
.
An explanation of each key:
:ref
: a name which can be used as a reference in the description (:desc
):desc
: a description of the option.:coerce
: coerce string to given type.:alias
: mapping of short name to long name.:default
: default value.:default-desc
: a string representation of the default value.:require
:true
make this opt required.:validate
: a function used to validate the value of this opt (as described in the Validate section).
Given the above spec
you can print options as follows:
(println (cli/format-opts {:spec spec :order [:from :to :paths :pretty]}))
This will print:
-i, --from <format> edn The input format. <format> can be edn, json or transit.
-o, --to <format> json The output format. <format> can be edn, json or transit.
--paths src test Paths of files to transform.
-p, --pretty Pretty-print output.
As options can often be re-used in multiple subcommands, you can determine the
order and selection of printed options with :order
. If you don't want to use
:order
and simply want to present the options as written, you can also use a
vector of vectors for the spec:
[[:pretty {:desc "Pretty-print output."
:alias :p}]
[:paths {:desc "Paths of files to transform."
:coerce []
:default ["src" "test"]
:default-desc "src test"}]]
To handle subcommands, use dispatch.
An example. Say we want to create a CLI that can be called as:
$ example copy <file> --dry-run
$ example delete <file> --recursive --depth 3
This can be accomplished by doing the following:
(ns example
(:require [babashka.cli :as cli]))
(defn copy [m]
(assoc m :fn :copy))
(defn delete [m]
(assoc m :fn :delete))
(defn help [m]
(assoc m :fn :help))
(def table
[{:cmds ["copy"] :fn copy :args->opts [:file]}
{:cmds ["delete"] :fn delete :args->opts [:file]}
{:cmds [] :fn help}])
(defn -main [& args]
(cli/dispatch table args {:coerce {:depth :long}}))
Calling the example
namespace's -main
function can be done using clojure -M -m example
or bb -m example
.
The last entry in the dispatch-table
always matches and calls the help function.
When running clj -M -m example --help
, dispatch
calls help
which returns:
{:opts {:help true}, :dispatch [], :fn :help}
When running clj -M -m example copy the-file --dry-run
, dispatch
calls copy
,
which returns:
{:cmds ["copy" "the-file"], :opts {:file "the-file" :dry-run true},
:dispatch ["copy"], :fn :copy}
When running clj -M -m example delete the-file --depth 3
, dispatch
calls delete
which returns:
{:cmds ["delete" "the-file"], :opts {:depth 3, :file "the-file"},
:dispatch ["delete"], :fn :delete}
See neil for a real world example of a CLI that uses subcommands.
Additional parse-arg
options may be passed in each table entry:
(def table
[{:cmds ["copy"] :fn copy :args->opts [:file] :alias {:f :file :restrict true}}
{:cmds ["delete"] :fn delete :args->opts [:file]}
{:cmds [] :fn help}])
For documentation on babashka tasks, go here.
To parse options to your tasks, add [babashka.cli :as cli]
to
:requires
. Then you can parse the options in :init
:
:init (def cmd-line-opts (cli/parse-opts *command-line-args*)))
and then use this in any task:
(when-not (:skip-bump cmd-line-opts)
(run 'bump-release))
and your tasks can then be called with options:
$ bb publish --skip-bump
You can control parsing behavior by adding :org.babashka/cli
metadata to
Clojure functions. It does not introduce a dependency on babashka.cli
itself. Not adding any metadata will result in string values, which in many
cases may already be a reasonable default.
Adding support for this library will cause less friction with shell usage,
especially on Windows since you need less quoting. You can support the same
function for both clojure -X
and clojure -M
style invocations without
writing extra boilerplate.
In your deps.edn
:aliases
entry, add:
:exec {:extra-deps {org.babashka/cli {:mvn/version "0.3.35"}}
:main-opts ["-m" "babashka.cli.exec"]}
Now you can call any function that accepts a map argument. E.g.:
$ clojure -M:exec clojure.core prn :a 1 :b 2
{:a "1", :b "2"}
Use :org.babashka/cli
metadata for coercions:
(ns my-ns)
(defn foo
{:org.babashka/cli {:coerce {:a :symbol
:b :long}}}
;; map argument:
[m]
;; print map argument:
(prn m))
$ clojure -M:exec my-ns foo :a foo/bar :b 2 :c vanilla
{:a foo/bar, :b 2, :c "vanilla"}
Note that any library can add support for babashka CLI without depending on babashka CLI.
An example that specializes babashka.cli
usage to a function:
:prn {:extra-deps {org.babashka/cli {:mvn/version "0.3.35"}}
:main-opts ["-m" "babashka.cli.exec" "clojure.core" "prn"]}
$ clojure -M:prn --foo=bar --baz
{:foo "bar" :baz true}
You can also pre-define the exec function in :exec-fn
:
:prn {:extra-deps {org.babashka/cli {:mvn/version "0.3.35"}}
:exec-fn clojure.core/prn
:main-opts ["-m" "babashka.cli.exec"]}
To alter the parsing behavior of functions you don't control, you can add
:org.babashka/cli
data in the deps.edn
alias:
:prn {:deps {org.babashka/cli {:mvn/version "0.3.35"}}
:exec-fn clojure.core/prn
:main-opts ["-m" "babashka.cli.exec"]
:org.babashka/cli {:coerce {:foo :long}}}
$ clojure -M:prn --foo=1
{:foo 1}
.clojure/deps.edn
alias:
:antq {:deps {org.babashka/cli {:mvn/version "0.3.35"}
com.github.liquidz/antq {:mvn/version "1.7.798"}}
:paths []
:main-opts ["-m" "babashka.cli.exec" "antq.tool" "outdated"]
:org.babashka/cli {:coerce {:skip []}}}
On the command line you can now run it with:
$ clj -M:antq --upgrade
Note that we are calling the same outdated
function that you normally call
with -T
:
$ clj -Tantq outdated :upgrade true
even though antq has its own -main
function.
Note that we added the :org.babashka/cli {:coerce {:skip []}}
data in the
alias to make sure that --skip
options get collected into a vector:
clj -M:antq --upgrade --skip github-action
vs.
clj -Tantq outdated :upgrade true :skip '["github-action"]'
The following projects have added support for babashka CLI. Feel free to add a PR to list your project as well!
In deps.edn
create an alias:
:codox {:extra-deps {org.babashka/cli {:mvn/version "0.3.35"}
codox/codox {:mvn/version "0.10.8"}}
:exec-fn codox.main/generate-docs
;; default arguments:
:exec-args {:source-paths ["src"]}
:org.babashka/cli {:coerce {:source-paths []
:doc-paths []
:themes [:keyword]}}
:main-opts ["-m" "babashka.cli.exec"]}
CLI invocation:
$ clojure -M:codox --output-path /tmp/out
In deps.edn
create an alias:
:kaocha {:extra-deps {org.babashka/cli {:mvn/version "0.3.35"}
lambdaisland/kaocha {:mvn/version "1.66.1034"}}
:exec-fn kaocha.runner/exec-fn
:exec-args {} ;; insert default arguments here
:org.babashka/cli {:alias {:watch :watch?
:fail-fast :fail-fast?}
:coerce {:skip-meta :keyword
:kaocha/reporter [:symbol]}}
:main-opts ["-m" "babashka.cli.exec"]}
Now you are able to use kaocha's exec-fn to be used as a CLI:
$ clj -M:kaocha --watch --fail-fast --kaocha/reporter kaocha.report/documentation
In deps.edn
create an alias:
:build {:deps {org.babashka/cli {:mvn/version "0.3.35"}
io.github.clojure/tools.build {:git/tag "v0.8.2" :git/sha "ba1a2bf"}}
:paths ["."]
:ns-default build
:main-opts ["-m" "babashka.cli.exec"]}
Now you can call your build functions as CLIs:
clj -M:build jar --verbose
In deps.edn
create an alias:
:graph {:deps {org.babashka/cli {:mvn/version "0.3.35"}
org.clojure/tools.deps.graph {:mvn/version "1.1.68"}}
:exec-fn clojure.tools.deps.graph/graph
:exec-args {} ;; insert default arguments here
:org.babashka/cli {:coerce {:trace-omit [:symbol]}}
:main-opts ["-m" "babashka.cli.exec"]}
Then invoke on the command line:
clj -M:graph --size --output graph.png
This tool can be used to run clojure exec functions with lein.
An example with clj-new
:
In ~/.lein/profiles.clj
put:
{:clj-1.11 {:dependencies [[org.clojure/clojure "1.11.1"]]}
:clj-new {:dependencies [[org.babashka/cli "0.3.35"]
[com.github.seancorfield/clj-new "1.2.381"]]}
:user {:aliases {"clj-new" ["with-profiles" "+clj-1.11,+clj-new"
"run" "-m" "babashka.cli.exec"
{:exec-args {:env {:description "My project"}}
:coerce {:verbose :long
:args []}
:alias {:f :force}}
"clj-new"]}}}
After that you can use lein clj-new app
to create a new app:
$ lein clj-new app --name foobar/baz --verbose 3 -f
Copyright © 2022 Michiel Borkent
Distributed under the MIT License. See LICENSE.