by networktocode

networktocode / ntc-templates

TextFSM templates for parsing show commands of network devices

503 Stars 361 Forks Last release: about 2 months ago (v1.5.0) Other 955 Commits 9 Releases

Available items

No Items, yet!

The developer of this repository has not created any items for sale yet. Need a bug fixed? Help with integration? A different license? Create a request here:

Build Status Code style: black


Repository of TextFSM Templates for Network Devices, and Python wrapper for TextFSM's CliTable.

TextFSM is a project built by Google that takes CLI string output and passes each line through a series of regular expressions until it finds a match. The regular expressions use named capture groups to build a text table out of the significant text. The names of the capture groups are used as column headers, and the captured values are stored as rows in the table.

This project provides a large collection of TextFSM Templates (text parsers) for a variety of Networking Vendors. In addition to the templates, there is a function that will convert the CLI output into a CliTable object; the resulting text table is converted into a list of dictionaries mapping the column headers with each row in the table.

Installation and Usage

The project can be installed using either Git or PyPI; if you would like to use the templates outside of this project, then Git is the recommended approach.


$ git clone [email protected]:networktocode/ntc-templates.git
# Optional steps to install ntc-templates as a python package
$ pip install -e ntc-templates/

The install can also include the required dev packages, which can be useful for adding or editing templates:

$ git clone [email protected]:networktocode/ntc-templates.git
# Optional steps to install ntc-templates as a python package
$ pip install -e ntc-templates/[dev]


$ pip install ntc_templates

To include the dev packages:

$ pip install ntc_templates[dev]


>>> from ntc_templates.parse import parse_output
>>> vlan_output = (
        "VLAN Name                             Status    Ports\n"
        "---- -------------------------------- --------- -------------------------------\n"
        "1    default                          active    Gi0/1\n"
        "10   Management                       active    \n"
        "50   VLan50                           active    Fa0/1, Fa0/2, Fa0/3, Fa0/4, Fa0/5,\n"
        "                                                Fa0/6, Fa0/7, Fa0/8\n"
>>> vlan_parsed = parse_output(platform="cisco_ios", command="show vlan", data=vlan_output)
>>> vlan_parsed
        'vlan_id': '1',
        'name': 'default',
        'status': 'active',
        'interfaces': ['Gi0/1']
        'vlan_id': '10',
        'name': 'Management',
        'status': 'active',
        'interfaces': []
        'vlan_id': '50',
        'name': 'VLan50', 'status': 'active',
        'interfaces': ['Fa0/1', 'Fa0/2', 'Fa0/3', 'Fa0/4', 'Fa0/5', 'Fa0/6', 'Fa0/7', 'Fa0/8']


Pull requests are welcomed and automatically built and tested through TravisCI.

New Templates

To contribute new templates, each new pull request must include the following:

  • TextFSM template
  • Modified version of the index file
  • Tests
    • Raw version of text to be parsed
    • YAML file containing the expected parsed dictionary

TextFSM Template

TextFSM templates should be placed in the

directory and should adhere to the following NTC-Templates style. The TextFSM template name should be in the following format:

The template should be named using:

{{ vendor_os }}_{{ command_with_underscores }}.textfsm

Ex: ciscoiosshowcdpneighbors.textfsm

Note: The vendor name must be valid from the os_choices tests, which is primarily based on Netmiko list of supported vendors. New vendors added should adhere to vendor_os.

Ex: vmware_nsx


The capture group names should be in UPPERCASE.

An example of the proper format is shown below.

Value TIME (\d+:\d+:\d+)
Value TIMEZONE (\S+)
Value DAYWEEK (\w+)
Value MONTH (\d+)
Value DAY (\d+)
Value YEAR (\d+)

Start ^${TIME}\s+${TIMEZONE}\s+${DAYWEEK}\s+${DAY}/${MONTH}/${YEAR} -> Record ^. -> Error


If the raw output has a heading, the

state should match on the column headings and then transition to another state that will match the device's output table with the capture groups. This helps ensure the regex patterns for the capture groups are attempting to match the correct information, and allows templates to easily add additional States for tables that have different headings. Example:

Raw Output

... omitted
Network Next Hop Metric LocPrf Weight Path
*> 4294967295 4294967295 65535 1000 1000 1000 i

Sample Template ``` Start

Checking for header

^\sNetwork\s+Next(?:\s+|-)[Hh]op\s+Metric\s+LocPrf\s+Weight\s+Path\s$$ -> BGPTable

BGPTable ... omitted ```

Each state should end with

^. -> Error
. This helps to ensure we're accounting for every line within the raw output for the command. This doesn't mean we have to capture all the data as a Value, but we do have to account for it. In addition, it is also good to provide an expression to match blank lines,

An example would be the following raw output:

NAME: "3640 chassis", DESCR: "3640 chassis"
PID: , VID: 0xFF, SN: FF1045C5

The template would be the following: ``` Value NAME (.) Value DESCRIPTION (.)

Start ^NAME:\s+"${NAME}",\sDESCR:\s+"${DESCRIPTION}" ^PID:\s,\sVID:\s\S+,\sSN:\s\S+ ^\s*$$ ^. -> Error ```

Taking a look at the example template above, you notice that we're using \s* and \s+. These are the preferred way to match space characters, and should be used over the literal space character. For example,

This is not preferred$$
  • \s* accounts for zero or more spaces (use when the output may or may not have a space between characters)
  • \s+ accounts for one or more spaces (use when output will have a space, but could have more than one space)

Index File

The Index file binds the templates to the commands being run. Special care has been taken on ordering, as there is potential for issues. e.g.

show ip route
picking up for
show ip router vrf 
. We have used a combination of ordering, as defined:
  • OS in alphabetical order
  • Template name in length order
  • When length is the same, use alphabetical order of command name
  • Keep space between OS's


Template, Hostname, Platform, Command

same os, same length, used alphabetical order of command name

arista_eos_show_mlag.textfsm, ., arista_eos, sh[[ow]] ml[[ag]] arista_eos_show_vlan.textfsm, ., arista_eos, sh[[ow]] vl[[an]]

os in alphabetical order and space between cisco_asa and arista_eos

cisco_asa_dir.textfsm, .*, cisco_asa, dir

same os, template name length different and space between cisco_asa and cisco_ios

cisco_ios_show_capability_feature_routing.textfsm, ., cisco_ios, sh[[ow]] cap[[ability]] f[[eature]] r[[outing]] cisco_ios_show_interface_transceiver.textfsm, ., cisco_ios, sh[[ow]] int[[erface]] trans[[ceiver]] cisco_ios_show_cdp_neighbors_detail.textfsm, .*, cisco_ios, sh[[ow]] c[[dp]] neig[[hbors]] det[[ail]]


Tests will be located in

with the following hierarchy: -
./tests/{{ vendor_os }}/{{ command_name }}/


{{ command_name }}
directory should include the
file that includes the raw output of the command to be parsed, and the
file of the returned structured data.
$ ls tests/cisco_ios/show_clock/
Raw version of input text

The raw text file should contain only the output of the CLI command to be parsed. It should not contain the CLI command itself.

An example of the proper format is shown below:

$ cat tests/cisco_ios/show_clock/cisco_ios_show_clock.raw
*18:57:38.347 UTC Mon Oct 19 2015
YAML file containing expected parsed dictionary

The parsed file should match the data that is returned from the

function discussed in the beginning. Dictionary keys should be in lowercase.

The parsed text file should be placed in a directory in the

directory with the same name as the template file but replace
file extension with
. The raw text file and the parsed text file should be in the same directory. ex. ./tests/ciscoios/showclock/

There are available helpers to create the parsed file in the correct format (See Development Helper Scripts below).

An example of the proper format is shown below: ```bash

$ cat ./tests/ciscoios/showclock/ciscoiosshow_clock.yml

parsed_sample: - time: "18:57:38.347" timezone: "UTC" dayweek: "Mon" month: "Oct" day: "19" year: "2015" $ ```


files are supported per directory, and are encouraged, as there are differences depending on version, length, etc. Additional test files and more real life data helps ensure backwards compatibility is maintained as each template is updated and merged into the repo.

All YAML files must adhere to the YAML standards defined in the

file in the root directory. Yamllint provides thorough documentation of their configuration settings here.
Development Helper Scripts

A cli utility is provided to assist with properly building the parsed files. This utility depends on some packages listed in the dev install requirements; see Install and Usage for directions on installing the dev requirements. All arguments that can be passed to the script are mutually exclusive (i.e. you can only pass one argument). The file can be made executable with the

chmod +x development_scripts.py
command. The arguments are:
  • -y
    : Takes the path to a YAML file and ensures that the file adheres to the .yamllint settings
  • -yd
    : Takes a glob path to a directory or directories that will ensure all files ending in
    adhere to the .yamllint settings
  • -c
    : Takes the path to a
    file, and generates the parsed data and saves the results adjacent to the
    file following the defined standards in .yamllint.
  • -cd
    : Takes a glob path to a directory or directories containing
    files, and creates the appropriate parsed files in the appropriate directory.


arguments are designed to allow developers to generate the expected parsed file how they want, and ensure that the formatting adheres to the defined standard for this project.


arguments use
to generate the parsed data; this means that you can use these arguments to auto-generate the test
file(s) for new templates; just be sure that the template's parsing behavior meets expectations. In order for the data to be parsed, the template must be placed in
and the
file must be updated to correctly point to the template file(s).
$ /development_scripts.py -yd tests/cisco_ios/show_mac-address-table
$ ls tests/arista_eos/show_version/
$ ./development_scripts.py -c tests/arista_eos/show_version/arista_eos_show_version.txt
$ ls tests/arista_eos/show_version/
arista_eos_show_version.raw   arista_eos_show_version.yml

Updating/Fixing Existing Templates

When either fixing a bug within a template or adding additional Values to be captured, additional test files should be added to ensure backwards compatibility and that the new data is being parsed correctly.

To add additional raw/parsed tests for a command: - Navigate to

./tests/{{ vendor_os }}/{{ command_name }}/
- Create new
files within the directory, preferrably with a name identifying why the data is unique: * Existing raw:
* New raw:
* Existing parsed:
* New parsed:


You can test your changes locally within your Git branch before submitting a PR. If you do not have tox already installed, you can do that using pip or your systems package manager. Tox should be ran inside the ntc-templates root directory. The tox file is configured to run against python3.6, so either python3.6 needs to be available, or the tox.ini file will need to be updated with an available Python version. ```bash $ tox GLOB sdist-make: /home/admin/ntc-templates/setup.py py36 inst-nodeps: /home/admin/ntc-templates/.tox/dist/ntctemplates-1.3.0.zip py36 installed: appdirs==1.4.3,atomicwrites==1.3.0,attrs==19.3.0,black==19.10b0,Click==7.0,future==0.18.2,importlib-metadata==0.23,more-itertools==7.2.0,ntc-templates==1.3.0,packaging==19.2,pathspec==0.6.0,pluggy==0.13.0,py==1.8.0,pyparsing==2.4.5,pytest==5.2.4,PyYAML==5.1.2,regex==2019.11.1,six==1.13.0,terminal==0.4.0,textfsm==1.1.0,toml==0.10.0,typed-ast==1.4.0,wcwidth==0.1.7,yamllint==1.18.0,zipp==0.6.0 py36 runtests: PYTHONHASHSEED='3677750645' py36 runtests: commands[0] | black ./ --diff --check All done! ✨ 🍰 ✨ 8 files would be left unchanged. py36 runtests: commands[1] | yamllint tests/ py36 runtests: commands[2] | pytest -vv ================================================================ test session starts ================================================================= platform linux -- Python 3.6.8, pytest-5.2.4, py-1.8.0, pluggy-0.13.0 -- /home/jmcgill/repos/ntc-templates/.tox/py36/bin/python3.6 cachedir: .pytestcache rootdir: /home/jmcgill/repos/ntc-templates collected 428 items

tests/testindexorder.py::testindexordering PASSED [ 0%] tests/teststructureddataagainstparsedreferencefiles.py::testrawdataagainstmock[tests/alcatelsros/oammac-ping/alcatelsrosoammac-ping.raw] PASSED [ 0%] tests/teststructureddataagainstparsedreferencefiles.py::testrawdataagainstmock[tests/alcatelsros/showserviceidbase/alcatelsrosshowserviceidbase.raw] PASSED [ 0%] tests/teststructureddataagainstparsedreferencefiles.py::testrawdataagainstmock[tests/alcatelsros/showrouterbgproutesvpn-ipv4/alcatelsrosshowrouterbgproutesvpn-ipv4.raw] PASSED [ 0%] tests/teststructureddataagainstparsedreferencefiles.py::testrawdataagainstmock[tests/brocadefastiron/showlldpneighbors/brocadefastironshowlldpneighbors.raw] PASSED [ 1%] ... tests/teststructureddataagainstparsedreferencefiles.py::testrawdataagainstmock[tests/cisconxos/showipinterfacebrief/cisconxosshowipinterfacebrief.raw] PASSED [ 99%] tests/testtestcasesexists.py::testverifyparsedandreferencedata_exists PASSED [100%]

================================================================ 428 passed in 43.84s ================================================================ ______________________________________________________________________ summary _______________________________________________________________________ py36: commands succeeded congratulations :) $ ```


For any questions or comments, please feel free to swing by the networktocode slack channel.

Sign up here


From an outsiders view, some design choices, requirements, and testing procedures can seem arbitrary. The following list of FAQ is intended to help provide context and better guide users and contributors of ntc-templates.

Why is there a requirement to use

in every template?

Ensuring that the textfsm template can account for every line is the only method to ensure that data was not accidentally missed. Take the following example. Initially we account for status to be:

Value STATUS (up|down)

Given the result of:

Interface                      Status         Protocol Description
Gi0/0/1                        admin down     down
Gi0/0/2                        up             up       ISP Connection
Gi0/0/3                        down           down

The output would miss the G0/0/1 interface, since the

admin down
not known. If this was a low percentage use case, it can go undetected, and result in incorrect information being returned. Instead, by ensuring that we fail fast, an
is raised and hopefully GitHub Issue is put in.

Then why isn't

used in all templates?

Initially the controls were not as strong, so many templates were put in until issues like the previous became an issue.

Does the project support requests for additional templates or additional data in an existing template?

We are no longer considering additional template requests at this time. The project has existed for over 5 years (initially within ntc-ansible) and nearly 200 template at this point any additional requests are essentially edge use cases. Meaning, for five years of usage, no one else has asked for this feature. There is a limited maintainers who primarily use their free time to maintain the project.

Are you open to adding maintainers to the project?

Yes, we would consider giving a proven member of the project and community maintainer rights. Please inquiry emailing [email protected]

I simply want to add my template to the project, I do not want to add all of these tests and controls, can I just do so?

Short answer no, from an outsiders point of view the contributor requirements may seem overly complex, however features added by engineers rarely come back to support them. The burden of support is on the maintainers and a certain level of quality assurance is required for that to happen. That includes updating the index file appropriately and adding proper raw and expected value files.

Why don't you grab all of the data in the template?

There is no intention for ntc-templates to become feature complete, some of the data is less interesting, or can be better understood from other commands. This is actually an area where the project chose to be loose, as we do not want to over-burden the contributor. If you feel that the additional data should be added, you are welcome to add the feature, but it would not be considered a bug, and thus not supported by the maintainers of the this project.

Why does the index order matter?

The "greediness" of the template match ensures that there longest matches first. For example, if

show ip ospf
was above
show ip ospf database
, the
show ip ospf
template would be used in both cases. The additional steps are because of general programmatic hygiene.

Will you accept my template if I create it?

In most cases, yes. However, there are a few edge cases. For example if requesting to add a

show cdp neighbors
when there is already a
show cdp neighbors details
template created. That is additional complexity added to the project with little value.

Why was my issue closed?

The most likely reasons are:

  • Did not follow the Issue creation template.
  • Did not provide the data required to act upon the request.
  • A prolonged time with no response.

What is meant that this is a parsing project, not a data modeling project?

The project intends to parse, meaning post processing is assumed in order to normalize the data. This project does not intend to solve that problem set. This is often noted in keys being different between the same command on multiple OS's. This was not intentional as at first there was not strict enforcement. That being said, there is no intention to retrofit this use case for the above stated reasons. This use case is best handled in post processing.

I have never submitted a Pull Request before, how do I do so?

This is outside the scope of this project, but this video should provide the instructions on how to do so.

Does this work on windows?

Based on this PR it should, however this is not a supported option. We are willing to take in qualified Pull Requests to have the feature, but have no intention of actively supporting.

Can you provide general guidance?

This is best handled via real time communication. Feel free to join our slack community (sign up information above) and reach out on the #networktocode channel. Please be aware of timezones, downtimes, and help is performed based on goodwill and timing, and not guaranteed.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.