mgmt/docs/function-guide.md

Function guide

Overview

The mgmt tool has built-in functions which add useful, reactive functionality to the language. This guide describes the different function API's that are available. It is meant to instruct developers on how to write new functions. Since mgmt and the core functions are written in golang, some prior golang knowledge is assumed.

Theory

Functions in mgmt are similar to functions in other languages, however they also have a reactive component. Our functions can produce events over time, and there are different ways to write functions. For some background on this design, please read the original article on the subject.

Native Functions

Native functions are functions which are implemented in the mgmt language itself. These are currently not available yet, but are coming soon. Stay tuned!

Simple Function API

Most functions should be implemented using the simple function API. This API allows you to implement simple, static, pure functions that don't require you to write much boilerplate code. They will be automatically re-evaluated as needed when their input values change. These will all be automatically made available as helper functions within mgmt templates, and are also available for use anywhere inside mgmt programs.

You'll need some basic knowledge of using the types library which is included with mgmt. This library lets you interact with the available types and values in the mgmt language. It is very easy to use, and should be fairly intuitive. Most of what you'll need to know can be inferred from looking at example code.

To implement a function, you'll need to create a file that imports the lang/funcs/simple/ module. It should probably get created in the correct directory inside of: lang/core/. The function should be implemented as a simple.Scaffold in our API. It is then registered with the engine during init(). An example explains it best:

Example

package simple

import (
	"context"
	"fmt"

	"github.com/purpleidea/mgmt/lang/funcs/simple"
	"github.com/purpleidea/mgmt/lang/types"
)

// you must register your functions in init when the program starts up
func init() {
	// Example function that squares an int and prints out answer as an str.

	simple.ModuleRegister(ModuleName, "talkingsquare", &simple.Scaffold{
		T: types.NewType("func(int) str"), // declare the signature
		F: func(ctx context.Context, input []types.Value) (types.Value, error) {
			i := input[0].Int() // get first arg as an int64
			// must return the above specified value
			return &types.StrValue{
				V: fmt.Sprintf("%d^2 is %d", i, i * i),
			}, nil // no serious errors occurred
		},
	})
}

This simple function accepts one int as input, and returns one str. Functions can have zero or more inputs, and must have exactly one output. You must be sure to use the types library correctly, since if you try and access an input which should not exist (eg: input[2], when there are only two that are expected), then you will cause a panic. If you have declared that a particular argument is an int but you try to read it with .Bool() you will also cause a panic. Lastly, make sure that you return a value in the correct type or you will also cause a panic!

If anything goes wrong, you can return an error, however this will cause the mgmt engine to shutdown. It should be seen as the equivalent to calling a panic(), however it is safer because it brings the engine down cleanly. Ideally, your functions should never need to error. You should never cause a real panic(), since this could have negative consequences to the system.

Example

package simple

import (
	"context"
	"fmt"

	"github.com/purpleidea/mgmt/lang/funcs/simple"
	"github.com/purpleidea/mgmt/lang/types"
)

func init() {
	// This is the actual definition of the `len` function.
	simple.Register("len", &simple.Scaffold{
		T: types.NewType("func(?1) int"), // contains a unification var
		C: simple.TypeMatch([]string{     // match on any of these sigs
			"func(str) int",
			"func([]?1) int",
			"func(map{?1: ?2}) int",
		}),
		// The implementation is left as an exercise for the reader.
		F: Len,
	})
}

Simple Polymorphic Function API

Most functions should be implemented using the simple function API. If they need to have multiple polymorphic forms under the same name, with each resultant type match needing to be paired to a different implementation, then you can use this API. This is useful for situations when the functions differ in output type only.

Function API

To implement a reactive function in mgmt it must satisfy the Func interface. Using the Simple Function API is preferable if it meets your needs. Most functions will be able to use that API. If you really need something more powerful, then you can use the regular function API. What follows are each of the method signatures and a description of each.

Info

Info() *interfaces.Info

This returns some information about the function. It is necessary so that the compiler can type check the code correctly, and know what optimizations can be performed. This is usually the first method which is called by the engine.

Example

func (obj *FooFunc) Info() *interfaces.Info {
	return &interfaces.Info{
		Pure: true,
		Sig:  types.NewType("func(a int) str"),
	}
}

Init

Init(init *interfaces.Init) error

This is called to initialize the function. If something goes wrong, it should return an error. It is passed a struct that contains all the important information and pointers that it might need to work with throughout its lifetime. As a result, it will need to save a copy to that pointer for future use in the other methods.

Example

// Init runs some startup code for this function.
func (obj *FooFunc) Init(init *interfaces.Init) error {
	obj.init = init
	return nil
}

Call

Call is run when you want to return a new value from the function. It takes the input arguments to the function.

Example

func (obj *FooFunc) Call(ctx context.Context, args []types.Value) (types.Value, error) {
	return &types.StrValue{ // Our type system "str" (string) value.
		V: strconv.FormatInt(args[0].Int(), 10), // a golang string
	}, nil
}

Stream

Stream(context.Context) error

Stream is where any evented work is done. This method is started by the function engine. It will run this function once. It should call the obj.init.Event() method when it believes the function engine should run Call() again.

Implementing this is not required if you don't have events.

If the ctx closes, you must shutdown as soon as possible.

Example

// Stream starts a mainloop and runs Event when it's time to Call() again.
func (obj *FooFunc) Stream(ctx context.Context) error {

	ticker := time.NewTicker(time.Duration(1) * time.Second)
	defer ticker.Stop()

	// streams must generate an initial event on startup
	// even though ticker will send one, we want to be faster to first event
	startChan := make(chan struct{}) // start signal
	close(startChan)                 // kick it off!

	for {
		select {
		case <-startChan:
			startChan = nil // disable

		case <-ticker.C: // received the timer event
			// pass

		case <-ctx.Done():
			return nil
		}

		if err := obj.init.Event(ctx); err != nil {
			return err
		}
	}
}

Further considerations

There is some additional information that any function author will need to know. Each issue is listed separately below!

Function struct

Each function will implement methods as pointer receivers on a function struct. The naming convention for resources is that they end with a Func suffix.

Example

type FooFunc struct {
	init *interfaces.Init

	// this space can be used if needed
}

Function registration

All functions must be registered with the engine so that they can be found. This also ensures they can be encoded and decoded. Make sure to include the following code snippet for this to work.

import "github.com/purpleidea/mgmt/lang/funcs"

func init() { // special golang method that runs once
	funcs.Register("foo", func() interfaces.Func { return &FooFunc{} })
}

Functions inside of built-in modules will need to use the ModuleRegister method instead.

// moduleName is already set to "math" by the math package. Do this in `init`.
funcs.ModuleRegister(moduleName, "cos", func() interfaces.Func {
	return &CosFunc{}
})

Composite functions

Composite functions are functions which import one or more existing functions. This is useful to prevent code duplication in higher level function scenarios. Unfortunately no further documentation about this subject has been written. To expand this section, please send a patch! Please contact us if you'd like to work on a function that uses this feature, or to add it to an existing one! We don't expect this functionality to be particularly useful or common, as it's probably easier and preferable to simply import common golang library code into multiple different functions instead.

Frequently asked questions

(Send your questions as a patch to this FAQ! I'll review it, merge it, and respond by commit with the answer.)

Can I use global variables?

Probably not. You must assume that multiple copies of your function may be used at the same time. If they require a global variable, it's likely this won't work. Instead it's probably better to use a struct local variable if you need to store some state.

There might be some rare instances where a global would be acceptable, but if you need one of these, you're probably already an internals expert. If you think they need to lock or synchronize so as to not overwhelm an external resource, then you have to be especially careful not to cause deadlocking the mgmt engine.

Can I write functions in a different language?

Currently golang is the only supported language for built-in functions. We might consider allowing external functions to be imported in the future. This will likely require a language that can expose a C-like API, such as python or ruby. Custom golang functions are already possible when using mgmt as a lib.

What new functions need writing?

There are still many ideas for new functions that haven't been written yet. If you'd like to contribute one, please contact us and tell us about your idea!

Can I generate many different FuncValue implementations from one function?

Yes, you can use a function generator in golang to build multiple different implementations from the same function generator. You just need to implement a function which returns a golang type of func([]types.Value) (types.Value, error) which is what FuncValue expects. The generator function can use any input it wants to build the individual functions, thus helping with code reuse.

How do I determine the signature of my simple, polymorphic function?

The determination of the input portion of the function signature can be determined by inspecting the length of the input, and the specific type each value has. Length is done in the standard golang way, and the type of each element can be ascertained with the Type() method available on every value.

Knowing the output type is trickier. If it can not be inferred in some manner, then the only way is to keep track of this yourself. You can use a function generator to build your FuncValue implementations, and pass in the unique signature to each one as you are building them. Using a generator is a common technique which was mentioned previously.

One obvious situation where this might occur is if your function doesn't take any inputs! An example math.fortytwo() function was implemented that demonstrates the use of function generators to pass the type signatures into the implementations.

Where can I find more information about mgmt?

Additional blog posts, videos and other material is available!.

Suggestions

If you have any ideas for API changes or other improvements to function writing, please let us know! We're still pre 1.0 and pre 0.1 and happy to break API in order to get it right!