Functions
Dagger Functions are regular code, written in a supported programming language, and running in containers. Dagger Functions let you encapsulate common operations or workflows into discrete units with clear inputs and outputs.
Here's an example of a simple Dagger Function:
from dagger import function, object_type
@object_type
class MyModule:
@function
def hello(self) -> str:
return "Hello, world"
Here is an example call for this Dagger Function:
dagger call hello
The result will be:
Hello, world
Here's an example of a more complex Dagger Function, which calls a remote API method:
from dagger import dag, function, object_type
@object_type
class MyModule:
@function
async def get_user(self) -> str:
return await (
dag.container()
.from_("alpine:latest")
.with_exec(["apk", "add", "curl"])
.with_exec(["apk", "add", "jq"])
.with_exec(
["sh", "-c", "curl https://randomuser.me/api/ | jq .results[0].name"]
)
.stdout()
)
Here is an example call for this Dagger Function:
dagger call get-user
The result will look something like this:
{
"title": "Mrs",
"first": "Beatrice",
"last": "Lavigne"
}
Class methods
Dagger Functions are implemented as @dagger.function decorated methods, of a @dagger.object_type decorated class.
It's possible for a module to implement multiple classes (object types), but the first one needs to have a name that matches the module's name, in PascalCase. This object is sometimes referred to as the main object.
For example, for a module initialized with dagger init --name=my-module
,
the main object needs to be named MyModule
.
Arguments
Dagger Functions, just like regular functions, can accept arguments. In addition to basic types (string, boolean, integer, arrays...), Dagger also defines powerful core types which functions can use for their arguments, such as Directory
, Container
, Service
, Secret
, and many more.
Here's an example of modifying the previous Dagger Function to accept a string argument:
from dagger import dag, function, object_type
@object_type
class MyModule:
@function
async def get_user(self, gender: str) -> str:
return await (
dag.container()
.from_("alpine:latest")
.with_exec(["apk", "add", "curl"])
.with_exec(["apk", "add", "jq"])
.with_exec(
[
"sh",
"-c",
f"curl https://randomuser.me/api/?gender={gender} | jq .results[0].name",
]
)
.stdout()
)
Here is an example call for this Dagger Function:
dagger call get-user --gender=male
The result will look something like this:
{
"title": "Mr",
"first": "Hans-Werner",
"last": "Thielen"
}
Here's an example of a Dagger Function that accepts a Directory
as argument:
import dagger
from dagger import dag, function, object_type
@object_type
class MyModule:
@function
async def tree(self, src: dagger.Directory, depth: str) -> str:
return await (
dag.container()
.from_("alpine:latest")
.with_mounted_directory("/mnt", src)
.with_workdir("/mnt")
.with_exec(["apk", "add", "tree"])
.with_exec(["tree", "-L", depth])
.stdout()
)
Here is an example call for this Dagger Function:
dagger call tree --src=. --depth=1
The result will look like this:
.
├── LICENSE
├── dagger
└── dagger.json
2 directories, 2 files
You can also pass a remote Git reference, and the Dagger CLI will convert it to a Directory
referencing the contents of that repository. Here is an example call that lists the source code for the Dagger CLI, from the main
branch of the Dagger GitHub repository:
dagger call tree --src=https://github.com/dagger/dagger#main:cmd/dagger --depth=1
The result will be the same file listing as this GitHub page:
.
├── call.go
├── cloud.go
├── debug.go
├── engine.go
├── exec_nonunix.go
├── exec_unix.go
├── flags.go
├── functions.go
├── gen.go
├── licenses.go
├── listen.go
├── log.go
├── main.go
├── module.go
├── module_test.go
├── query.go
├── run.go
├── session.go
├── shell.go
└── version.go
1 directory, 20 files
When calling a Dagger Function from the CLI, its arguments are exposed as command-line flags. How the flag is interpreted depends on the argument type.
Optional arguments
Function arguments can be marked as optional. In this case, the Dagger CLI will not display an error if the argument is omitted in the function call.
Here's an example of a Dagger Function with an optional argument:
from dagger import function, object_type
@object_type
class MyModule:
@function
def hello(self, name: str | None) -> str:
if name is None:
name = "world"
return f"Hello, {name}"
Here is an example call for this Dagger Function, with the optional argument:
dagger call hello --name=John
The result will look like this:
Hello, John
Here is an example call for this Dagger Function, without the optional argument:
dagger call hello
The result will look like this:
Hello, world
The API considers any type that accepts a None
value to be optional. Thus,
if the name
argument is omitted from an API request, it will be set to None
automatically.
However, when used directly from Python code, it's in fact a required argument
since it has no default value. In this context, the term "optional" is the same
concept as the older typing.Optional[X]
annotation for a type that allows the None
value.
Default values
Function arguments can define a default value if no value is supplied for them.
Here's an example of a Dagger Function with a default value for an argument:
from dagger import function, object_type
@object_type
class MyModule:
@function
def hello(self, name: str = "world") -> str:
return f"Hello, {name}"
Here is an example call for this Dagger Function, without the required argument:
dagger call hello
The result will look like this:
Hello, world
For consistency between Python and the API, choose to always set optionals with a default value, even if it's:
name: str | None = None
The above is in fact necessary if the default value needs to be set from the function body instead of the signature (dynamic, complex or a mutable value).
Return values
Not only can Dagger Functions use Dagger's core types in their arguments; they can use them in their return value as well.
This opens powerful applications to Dagger Functions. For example, a Dagger Function that builds binaries could take a directory as argument (the source code) and return another directory (containing binaries) or a container image (with the binaries included).
Here's an example of a Dagger Function that accepts a Directory
containing a Go application's source code as input, compiles it into a binary, and returns a Container
with the binary:
import dagger
from dagger import dag, function, object_type
@object_type
class MyModule:
@function
def build(self, source: dagger.Directory, arch: str, os: str) -> dagger.Container:
dir_ = (
dag.container()
.from_("golang:1.21")
.with_mounted_directory("/src", source)
.with_workdir("/src")
.with_env_variable("GOARCH", arch)
.with_env_variable("GOOS", os)
.with_env_variable("CGO_ENABLED", "0")
.with_exec(["go", "build", "-o", "build/"])
.directory("/src/build")
)
return (
dag.container()
.from_("alpine:latest")
.with_directory("/usr/local/bin", dir_)
)
Here is an example call for this Dagger Function:
dagger call build --source=https://github.com/golang/example/#master:hello --os=linux --arch=amd64 terminal
This example chains two functions calls:
- a call to
Build()
, which builds a Go application from a remote GitHub repository and returns aContainer
with the compiled binary; - a call to
Terminal()
, which opens an interactive terminal session with theContainer
returned by the previous function.
The result will be an interactive terminal session with the built container, which you can use to validate the compiled Go binary:
/ # cd /usr/local/bin
/usr/local/bin # ls
hello
/usr/local/bin # ./hello
Hello, world!
/usr/local/bin #
Type annotations
Even though the Python runtime doesn’t enforce type annotations at runtime, it’s important to define them with Dagger Functions. The Python SDK needs the typing information at runtime to correctly report to the API. It can’t rely on type inference, which is only possible for external static type checkers.
If a function doesn’t have a return type annotation, it’ll be declared as None
,
which translates to the dagger.Void type in the API:
@function
def hello(self):
return "Hello world!"
# Error: cannot convert string to Void
It’s fine however, when no value actually needs to be returned:
@function
def hello(self):
...
# no return