script

README

script

script is a collection of utilities for doing the kind of tasks that shell scripts are good at: reading files, counting lines, matching strings, and so on. Why shouldn't it be as easy to write system administration programs in Go as it is in a typical shell? script aims to make it just that easy.

What can I do with it?

Let's see a simple example. Suppose you want to read the contents of a file as a string:

contents, err := script.File("test.txt").String()

That looks straightforward enough, but suppose you now want to count the lines in that file.

numLines, err := script.File("test.txt").CountLines()

For something a bit more challenging, let's try counting the number of lines in the file which match the string "Error":

numErrors, err := script.File("test.txt").Match("Error").CountLines()

Wait, what?

Those chained function calls look a bit weird. What's going on there?

One of the neat things about the Unix shell, and its many imitators, is the way you can compose operations into a pipeline:

cat test.txt | grep Error | wc -l

The output from each stage of the pipeline feeds into the next, and you can think of each stage as a filter which passes on only certain parts of its input to its output.

By comparison, writing shell-like scripts in raw Go is much less convenient, because everything you do returns a different data type, and you must (or at least should) check errors following every operation.

In scripts for system administration we often want to compose different operations like this in a quick and convenient way. If an error occurs somewhere along the pipeline, we would like to check this just once at the end, rather than after every operation.

Everything is a pipe

The script library allows us to do this because everything is a pipe (specifically, a script.Pipe). To create a pipe, start with a source like File():

var p script.Pipe
p = script.File("test.txt")

You might expect File() to return an error if there is a problem opening the file, but it doesn't. We will want to call a chain of methods on the result of File(), and it's inconvenient to do that if it also returns an error.

Instead, you can check the error status of the pipe at any time by calling its Error() method:

p = script.File("test.txt")
if p.Error() != nil {
    log.Fatalf("oh no: %v", p.Error())
}

What use is a pipe?

Now, what can you do with this pipe? You can call a method on it:

var q script.Pipe
q = p.Match("Error")

Note that the result of calling a method on a pipe is another pipe. You can do this in one step, for convenience:

var q script.Pipe
q = script.File("test.txt").Match("Error")

Handling errors

Woah, woah! Just a minute! What if there was an error opening the file in the first place? Won't Match blow up if it tries to read from a non-existent file?

No, it won't. As soon as an error status is set on a pipe, all operations on the pipe become no-ops. Any operation which would normally return a new pipe just returns the old pipe unchanged. So you can run as long a pipeline as you want to, and if an error occurs at any stage, nothing will crash, and you can check the error status of the pipe at the end.

(Seasoned Gophers will recognise this as the errWriter pattern described by Rob Pike in the blog post Errors are values.)

Getting output

A pipe is useless if we can't get some output from it. To do this, you can use a sink, such as String():

result, err := q.String()
if err != nil {
    log.Fatalf("oh no: %v", err)
}
fmt.Println(result)

Errors

Note that sinks return an error value in addition to the data. This is the same value you would get by calling p.Error(). If the pipe had an error in any operation along the pipeline, the pipe's error status will be set, and a sink operation which gets output will return the zero value, plus the error.

numLines, err := script.File("doesnt_exist.txt").CountLines()
fmt.Println(numLines)
// Output: 0
if err != nil {
	    log.Fatal(err)
}
// Output: open doesnt_exist.txt: no such file or directory

CountLines() is another useful sink, which simply returns the number of lines read from the pipe.

Closing pipes

If you've dealt with files in Go before, you'll know that you need to close the file once you've finished with it. Otherwise, the program will retain what's called a file handle (the kernel data structure which represents an open file). There is a limit to the total number of open file handles for a given program, and for the system as a whole, so a program which leaks file handles will eventually crash, and will waste resources in the meantime.

How does script handle this? Simple. Any operation which completely reads the data from a pipe (such as Match() or String()) also closes the pipe by calling p.Close(). This closes the original data source that was used to create the pipe.

If the pipe was created from something that doesn't need to be closed, such as a string, then calling Close() simply does nothing.

TL;DR you don't need to worry about closing pipes; script will do it for you.

Sources, filters, and sinks

So script provides three types of pipe operations:

1. Sources create pipes from input in some way (for example, File() opens a file).
2. Filters read from a pipe and filter the data in some way (for example Match() passes on only lines which contain a given string).
3. Sinks get the output from a pipeline in some useful form (for example String() returns the contents of the pipe as a string), along with any error status.

Let's look at the source, filter, and sink options that script provides.

Sources

These are operations which create a pipe.

File

File() creates a pipe that reads from a file.

p = script.File("test.txt")
output, err := p.String()
fmt.Println(output)
// Output: contents of file

Echo

Echo() creates a pipe containing a given string:

p := script.Echo("Hello, world!")
output, err := p.String()
fmt.Println(output)
// Output: Hello, world!

Exec

Exec() runs a given command and creates a pipe containing its combined output (stdout and stderr). If there was an error running the command, the pipe's error status will be set.

p := script.Exec("echo hello")
output, err := p.String()
fmt.Println(output)
// Output: hello

Note that Exec() can also be used as a filter, in which case the given command will read from the pipe as its standard input.

Exit status

If the command returns a non-zero exit status, the pipe's error status will be set to the string "exit status X", where X is the integer exit status.

p := script.Exec("ls doesntexist")
output, err := p.String()
fmt.Println(err)
// Output: exit status 1

For convenience, you can get this value directly as an integer by calling ExitStatus() on the pipe:

p := script.Exec("ls doesntexist")
var exit int = p.ExitStatus()
fmt.Println(exit)
// Output: 1

The value of ExitStatus() will be zero unless the pipe's error status matches the string "exit status X", where X is a non-zero integer.

Error output

Even in the event of a non-zero exit status, the command's output will still be available in the pipe. This is often helpful for debugging. However, because String() is a no-op if the pipe's error status is set, if you want output you will need to reset the error status before calling String():

p := Exec("man bogus")
p.SetError(nil)
output, err := p.String()
fmt.Println(output)
// Output: No manual entry for bogus

Stdin

Stdin() creates a pipe which reads from the program's standard input.

p := script.Stdin()
output, err := p.String()
fmt.Println(output)
// Output: [contents of standard input]

Filters

Filters are operations on an existing pipe that also return a pipe, allowing you to chain filters indefinitely.

Match

Match() returns a pipe containing only the input lines which match the supplied string:

p := script.File("test.txt").Match("Error")

MatchRegexp

MatchRegexp() is like Match(), but takes a compiled regular expression instead of a string.

p := script.File("test.txt").MatchRegexp(regexp.MustCompile(`E.*r`))

Reject

Reject() is the inverse of Match(). Its pipe produces only lines which don't contain the given string:

p := script.File("test.txt").Match("Error").Reject("false alarm")

RejectRegexp

RejectRegexp() is like Reject(), but takes a compiled regular expression instead of a string.

p := script.File("test.txt").Match("Error").RejectRegexp(regexp.MustCompile(`false|bogus`))

EachLine

EachLine() lets you create custom filters. You provide a function, and it will be called once for each line of input. If you want to produce output, your function can write to a supplied strings.Builder. The return value from EachLine is a pipe containing your output.

p := script.File("test.txt")
q := p.EachLine(func(line string, out *strings.Builder) {
	out.WriteString("> " + line + "\n")
})
output, err := q.String()
fmt.Println(output)

Exec

Exec() runs a given command, which will read from the pipe as its standard input, and returns a pipe containing the command's combined output (stdout and stderr). If there was an error running the command, the pipe's error status will be set.

Apart from connecting the pipe to the command's standard input, the behaviour of an Exec() filter is the same as that of an Exec() source.

// `cat` copies its standard input to its standard output.
p := script.Echo("hello world").Exec("cat")
output, err := p.String()
fmt.Println(output)
// Output: hello world

Sinks

Sinks are operations which return some data from a pipe, ending the pipeline.

String

The simplest sink is String(), which just returns the contents of the pipe as a string, plus an error:

contents, err := script.File("test.txt").String()

Note that String(), like all sinks, consumes the complete output of the pipe, closing the input reader if appropriate. Therefore, calling String() (or any other sink method) again on the same pipe will return an error:

p := script.File("test.txt")
_, _ = p.String()
_, err := p.String()
fmt.Println(err)
// Output: read test.txt: file already closed

CountLines

CountLines(), as the name suggests, counts lines in its input, and returns the number of lines as an integer, plus an error:

numLines, err := script.File("test.txt").CountLines()

WriteFile

WriteFile() writes the contents of the pipe to a named file. It returns the number of bytes written, or an error:

wrote, err := script.File("source.txt").WriteFile("destination.txt")

AppendFile

AppendFile() is like WriteFile(), but appends to the destination file instead of overwriting it. It returns the number of bytes written, or an error:

wrote, err := script.Echo("Got this far!").AppendFile("logfile.txt")

Stdout

Stdout() writes the contents of the pipe to the program's standard output. It returns the number of bytes written, or an error:

p := Echo("hello world")
wrote, err := p.Stdout()

In conjunction with Stdin(), Stdout() is useful for writing programs which filter input. For example, here is a simple implementation of the Unix cat utility (which just copies its input to its output):

package main

import "github.com/bitfield/script"

func main() {
	script.Stdin().Stdout()
}

Writing your own pipe operations

There's nothing to stop you writing your own sources, sinks, or filters (in fact, that would be excellent. Please submit a pull request if you want to add them to the standard operations supplied with script.)

Writing a source

All a pipe source has to do is return a pointer to a script.Pipe. To be useful, a pipe needs to have a Reader attached to it. This is simply anything that implements io.Reader.

Echo() is a simple example, which just creates a pipe containing a string:

func Echo(s string) *script.Pipe {
	return script.NewPipe().WithReader(strings.NewReader(s))
}

Let's break this down:

• We create a strings.Reader to be our data source, using strings.NewReader on the supplied string.
• We create a new pipe with NewPipe().
• We attach the reader to the pipe with WithReader().

Simple, right? One more thing: when you create a pipe from a data source that needs to be closed after reading (such as a file), use WithCloser() instead of WithReader().

Here's an implementation of File(), for example:

func File(name string) *script.Pipe {
	r, err := os.Open(name)
	if err != nil {
		return script.NewPipe().WithError(err)
	}
	return script.NewPipe().WithCloser(r)
}

You can also see from this example how to create a pipe with error status:

return script.NewPipe().WithError(err)

Writing a filter

Filters are methods on pipes, that return pipes. For example, here's a simple filter which just reads and rejects all input, returning an empty pipe:

func (p *script.Pipe) RejectEverything() *script.Pipe {
	if p.Error() != nil {
		return &p
	}
	defer p.Close()
	_, err := ioutil.ReadAll(p.Reader)
	if err != nil {
		p.SetError(err)
		return &p
	}
	return script.Echo("")
}

Important things to note here:

• The first thing we do is check the pipe's error status. If this is set, we do nothing, and just return the original pipe.
• We close the source pipe once we successfully read all data from it.
• If an error occurs, we set the pipe's error status, using p.SetError(), and return the pipe.

Filters must not log anything, terminate the program, or return anything but *script.Pipe.

As you can see from the example, the pipe's reader is available to you as p.Reader. You can do anything with that that you can with an io.Reader.

If your method modifies the pipe (for example if it can set an error on the pipe), it must take a pointer receiver, as in this example. Otherwise, it can take a value receiver.

Writing a sink

Any method on a pipe which returns something other than a pipe is a sink. For example, here's an implementation of String():

func (p *script.Pipe) String() (string, error) {
	if p.Error() != nil {
		return "", p.Error()
	}
	defer p.Close()
	res, err := ioutil.ReadAll(p.Reader)
	if err != nil {
		p.SetError(err)
		return "", err
	}
	return string(res), nil
}

Again, the first thing we do is check the error status on the pipe. If it's set, we return the zero value (empty string in this case) and the error.

We then defer closing the pipe, and try to read from it. If we get an error reading the pipe, we set the pipe's error status and return the zero value and the error.

Otherwise, we return the result of reading the pipe, and a nil error.

Ideas

These are some ideas I'm playing with for additional features. If you feel like working on one of them, send a pull request. If you have ideas for other features, open an issue (or, better, a pull request).

Sources

• Get() makes a web request, like curl, and pipes the result
• Net() makes a network connection to a specified address and port, and reads the connection until it's closed
• ListFiles() takes a filesystem path or glob, and pipes the list of matching files
• Find() pipes a list of files matching various criteria (name, modified time, and so on)
• Processes() pipes the list of running processes, like ps.

Filters

• Ideas welcome!

Sinks

• Ideas equally welcome!

Examples

Since script is designed to help you write system administration programs, a few simple examples of such programs are included in the examples directory:

• cat
• grep

More examples would be welcome!

Use cases

The best libraries are designed to satisfy real use cases. If you have a sysadmin task which you'd like to implement with script, let me know by opening an issue.

Documentation

Overview

Package script aims to make it easy to write shell-type scripts in Go, for general system administration purposes: reading files, counting lines, matching strings, and so on.

Index

• type Pipe
  • func Echo(s string) *Pipe
  • func Exec(s string) *Pipe
  • func File(name string) *Pipe
  • func NewPipe() *Pipe
  • func Stdin() *Pipe
  • func (p *Pipe) AppendFile(fileName string) (int64, error)
  • func (p *Pipe) Close() error
  • func (p *Pipe) CountLines() (int, error)
  • func (p *Pipe) EachLine(process func(string, *strings.Builder)) *Pipe
  • func (p *Pipe) Error() error
  • func (p *Pipe) Exec(s string) *Pipe
  • func (p *Pipe) ExitStatus() int
  • func (p *Pipe) Match(s string) *Pipe
  • func (p *Pipe) MatchRegexp(re *regexp.Regexp) *Pipe
  • func (p *Pipe) Reject(s string) *Pipe
  • func (p *Pipe) RejectRegexp(re *regexp.Regexp) *Pipe
  • func (p *Pipe) SetError(err error)
  • func (p *Pipe) Stdout() (int, error)
  • func (p *Pipe) String() (string, error)
  • func (p *Pipe) WithCloser(r io.ReadCloser) *Pipe
  • func (p *Pipe) WithError(err error) *Pipe
  • func (p *Pipe) WithReader(r io.Reader) *Pipe
  • func (p *Pipe) WriteFile(fileName string) (int64, error)

Types

type Pipe

type Pipe struct {
	Reader io.ReadCloser
	// contains filtered or unexported fields
}

Pipe represents a pipe object with an associated Reader.

func Echo

func Echo(s string) *Pipe

Echo returns a pipe containing the supplied string.

func Exec

func Exec(s string) *Pipe

Exec runs an external command and returns a pipe containing the output. If the command had a non-zero exit status, the pipe's error status will also be set to the string "exit status X", where X is the integer exit status.

func File

func File(name string) *Pipe

File returns a *Pipe associated with the specified file. This is useful for starting pipelines. If there is an error opening the file, the pipe's error status will be set.

func NewPipe

func NewPipe() *Pipe

NewPipe returns a pointer to a new empty pipe.

func Stdin

func Stdin() *Pipe

Stdin returns a pipe which reads from the program's standard input.

func (*Pipe) AppendFile

func (p *Pipe) AppendFile(fileName string) (int64, error)

AppendFile appends the contents of the Pipe to the specified file, and closes the pipe after reading. It returns the number of bytes successfully written, or an error. If there is an error reading or writing, the pipe's error status is also set.

func (*Pipe) Close

func (p *Pipe) Close() error

Close closes the pipe's associated reader. This is always safe to do, because pipes created from a non-closable source will have an `ioutil.NopCloser` to call.

func (*Pipe) CountLines

func (p *Pipe) CountLines() (int, error)

CountLines counts lines from the pipe's reader, and returns the integer result, or an error. If there is an error reading the pipe, the pipe's error status is also set.

func (*Pipe) EachLine

func (p *Pipe) EachLine(process func(string, *strings.Builder)) *Pipe

EachLine calls the specified function for each line of input, passing it the line as a string, and a *strings.Builder to write its output to. The return value from EachLine is a pipe containing the contents of the strings.Builder.

func (*Pipe) Error

func (p *Pipe) Error() error

Error returns the last error returned by any pipe operation, or nil otherwise.

func (*Pipe) Exec

func (p *Pipe) Exec(s string) *Pipe

Exec runs an external command and returns a pipe containing the output. If the command had a non-zero exit status, the pipe's error status will also be set to the string "exit status X", where X is the integer exit status.

func (*Pipe) ExitStatus

func (p *Pipe) ExitStatus() int

ExitStatus returns the integer exit status of a previous command, if the pipe's error status is set, and if the error matches the pattern "exit status %d". Otherwise, it returns zero.

func (*Pipe) Match

func (p *Pipe) Match(s string) *Pipe

Match reads from the pipe, and returns a new pipe containing only lines which contain the specified string. If there is an error reading the pipe, the pipe's error status is also set.

func (*Pipe) MatchRegexp

func (p *Pipe) MatchRegexp(re *regexp.Regexp) *Pipe

MatchRegexp reads from the pipe, and returns a new pipe containing only lines which match the specified compiled regular expression. If there is an error reading the pipe, the pipe's error status is also set.

func (*Pipe) Reject

func (p *Pipe) Reject(s string) *Pipe

Reject reads from the pipe, and returns a new pipe containing only lines which do not contain the specified string. If there is an error reading the pipe, the pipe's error status is also set.

func (*Pipe) RejectRegexp

func (p *Pipe) RejectRegexp(re *regexp.Regexp) *Pipe

RejectRegexp reads from the pipe, and returns a new pipe containing only lines which don't match the specified compiled regular expression. If there is an error reading the pipe, the pipe's error status is also set.

func (*Pipe) SetError

func (p *Pipe) SetError(err error)

SetError sets the pipe's error status to the specified error.

func (*Pipe) Stdout

func (p *Pipe) Stdout() (int, error)

Stdout writes the contents of the pipe to the program's standard output. It returns the number of bytes successfully written, plus a non-nil error if the write failed or if there was an error reading from the pipe. If the pipe has error status, Stdout returns zero plus the existing error.

func (*Pipe) String

func (p *Pipe) String() (string, error)

String returns the contents of the Pipe as a string, or an error, and closes the pipe after reading. If there is an error reading, the pipe's error status is also set.

func (*Pipe) WithCloser

func (p *Pipe) WithCloser(r io.ReadCloser) *Pipe

WithCloser takes an io.ReadCloser and associates the pipe with that source.

func (*Pipe) WithError

func (p *Pipe) WithError(err error) *Pipe

WithError sets the pipe's error status to the specified error and returns the modified pipe.

func (*Pipe) WithReader

func (p *Pipe) WithReader(r io.Reader) *Pipe

WithReader takes an io.Reader which does not need to be closed after reading, and associates the pipe with that reader.

func (*Pipe) WriteFile

func (p *Pipe) WriteFile(fileName string) (int64, error)

WriteFile writes the contents of the Pipe to the specified file, and closes the pipe after reading. It returns the number of bytes successfully written, or an error. If there is an error reading or writing, the pipe's error status is also set.