TextFSMGo

TextFSMGo is just another but faster implementation of Google TextFSM, a template-based state machine for parsing semi-formatted text.

TextFSMGo is compatible with the existing TextFSM template (with some caveats): to write a template file check the original project documentation.

Build and install

To build and install TextFSMGo is it possible to leverage the Makefile recipe.

To build the project from the root directory of this repository run the command:

make build

it will build and store the executable in the ./dist directory of the repo.

To build and install TextFSMGo launch the command:

make install

Usage

Using the CLI tool

TextFSMGo provides a CLI tool allowing to parse some raw text using a textfsm template and producing a json output.

For example the following command parse the ip_cmd.raw file via the ip_cmd.textfsm template and prints the result in json format:

textfsmgo ./examples/data/ip_cmd.raw ./examples/data/ip_cmd.textfsm

The -o argument can be used to write the output to a local file:

textfsmgo -o result.json ./examples/data/ip_cmd.raw ./examples/data/ip_cmd.textfsm

It is possible to indent the json output by providing the -i argument.

Using the library

To use TextFSMGo declare it as dependency of your project

go get github.com/claudiolor/textfsmgo

To use TextFSMGo the first thing to do is writing the template) for the semi-formatted text to parse. Then it will be possible to create a state machine starting from a template file:

tmpl_file := "/path/to/template.textfsm"
text_to_parse := "..."
parser, err := textfsmgo.NewTextFSMParser(tmpl_file)
if err != nil {
    handleError(err)
}

res, err := parser.ParseTextToDicts(text_to_parse)
if err != nil {
    handleError(err, 1)
}

The parser.ParseTextToDicts() returns a slice of maps having string as keys and interface{} as value, which could actually be a string or a slice of strings.

Handling the result

Let's imagine we would like to parse the command ip a. The first thing to do is writing a template to parse it. At that point, we can feed the template to TextFSMGo so that it can parse the command output. TextFSMGo returns a slice of maps having string as keys and interface{} as values, which could be converted in a slice of structs, as in the example below:

type NetworkIf struct {
    ifname string
    macaddr string
    addresses []string
    mtu int
    state string
}

...
func main(){
    ...
    // Parse the text with TextFSMGo
    res, err := parser.ParseTextToDicts(text_to_parse)
    if err != nil {
        handleError(err, 1)
    }

    // Store the content of the Map into a slice of structs
    addresses_list := []NetworkIf{}
    for _, entry := range res {
        mtu, _ := strconv.Atoi(entry["mtu"].(string))
        addresses_list = append(
            addresses_list,
            NetworkIf{
                ifname:    entry["ifname"].(string),
                macaddr:   entry["macaddr"].(string),
                addresses: entry["addresses"].([]string),
                state:     entry["state"].(string),
                mtu:       mtu,
            },
        )
    }
}

JSON encoding

TextFSMGo provides an utility function that allows to encode the parsed result in json: ConvertResToJson(map_res *[]map[string]interface{}, indent bool). In the example below the result of the parsing is converted to json:

import "github.com/claudiolor/textfsmgo/pkg/textfsmg"

...

func main(){
    ...
    intend = false
    // Parse the text with TextFSMGo
    res, err := parser.ParseTextToDicts(text_to_parse)
    if err != nil {
        handleError(err, 1)
    }
    ...
    jsonRes, err := utils.ConvertResToJson(&res, intend)
}

A complete example can be found in the examples directory.

Performance

TextFSMGo, also due to the used programming language, guarantes a good level of performance. Some tests were performed parsing some raw data and encoding it in json and comparing the time required by TextFSMGo and original Python implementation. The tests were repeated increasing the size of the raw data. On average TextFSMGo was the 70% faster than the original Python implementation:

Tests were performed using a laptop with i5-1145G7 and 32GB of ram

Caveats

---

⚠️ Warning: I wrote TextFSMGo just for fun and to practice with Golang, so please consider it as is.

---

At the moment TextFSMGo has the following caveats:

• Named match groups in values definition are currently not supported, so values cannot contain dictionaries;
• Perl syntax of regex is not supported.