Skip to main content
Join the official 2020 Python Developers SurveyStart the survey!

Plot MAD output (and more).

Project description

This project aims to facilitate working with MADX from within Python. It contains the following major components:

  • MADX API: Build, parse and run MADX scripts.
  • Plot API: Plot MADX output in various formats.
  • Utilities: Convert MADX output tables to pandas data frames.

Script templating via Jinja is also supported.


The MADX API consists of three parts: building, parsing and running MADX scripts.


The builder API can be used for creating MADX scripts. The following example code shows the various features.

from madplot.madx.builder import Script

# At first generate a new script.
s = Script()

# Labeled or declaration statements can be created via `[]` access.
# This produces the following statement in the resulting MADX script:
# L = 5;
# N = 10;
s['L'] = 5
s['N'] = 10

# MADX commands can be created by accessing them through the script instance.
# Output: `DP: SBEND, L = L/2, ANGLE = 2*PI/(2*N);`.
s['DP'] = s.SBEND(L='L/2', ANGLE='2*PI/(2*N)')

# Output: `QF: MULTIPOLE, KNL = {0, 1/f};`.
s['QF'] = s.MULTIPOLE(KNL=[0, '1/f'])

# Sequences can be generated using the `Sequence` class.
from madplot.madx.builder import Sequence

with Sequence(refer='entry', l='N*L') as seq:
    for n in range(s.N):  # Python loop over number of cells.
        # Unlabeled statements can be just added the script instance.
        # Stored element definitions can be reused via attribute access of the script instance.
        # This produces the following output: `QF, at = 0 * L;`.
        seq += s.QF(at=f'{n} * L')

        # [...] Add more elements.

# Adding a sequence to the script will auto-expand it when dumping the script.
# This produces the following output:
# `LATTICE: sequence, refer = entry, l = N*L;`
# `    QF, at = 0 * L;`
# `    [...]`
# `endsequence;`
s['LATTICE'] = seq

# A script can be dumped by converting to `str`.
with open('example.seq', 'w') as f:

Complete code example

The following is a complete code example.

from madplot.madx.builder import Sequence, Script

s = Script()

s['N_cells'] = 60
s['L_cell'] = 13.45
s['f'] = 7.570366

s['DP'] = s.SBEND(L='L_cell/2', ANGLE='2*PI / (2*N_cells)')
s['QF'] = s.MULTIPOLE(KNL=['0', '1/f'])
s['QD'] = s.MULTIPOLE(KNL=['0', '-1/f'])

with Sequence(refer='entry', l='N_cells*L_cell') as seq:
    for n in range(s.N_cells):
        seq += s.QF(at=f'{n} * L_cell')
        seq += s.DP(at=f'{n} * L_cell')
        seq += s.QD(at=f'{n} * L_cell + 0.50 * L_cell')
        seq += s.DP(at=f'{n} * L_cell + 0.50 * L_cell')

s['FODO_LATTICE'] = seq

with open('example.seq', 'w') as f:

Advanced control

The following operations allow for advanced control statements.

  • Comments can be placed as strings: s += '// Comment'.
  • Re-evaluated (deferred) expressions (:=) can be created via the E class: from madplot.madx.builder import E; s += s.ealign(dx=E('ranf()')).
  • Any MADX command can be accessed via the script instance: s += s.TWISS(file='optics').


The parser.Parser class has two methods available:

  • Parser.raw_parse: This method parses the given script into its statements and returns a list thereof. The different statement types can be found in Parser._types. The literal values of command attributes will be returned.
  • Parser.parse: Parses the script into its statements as well but only returns non-comment non-variable declaration statements and interpolates any command attribute values.

For example:

>>> madx = '''
...     L = 5;
...     QF: QUADRUPOLE, k1 := pi/5, l = L;
... '''
>>> Parser.raw_parse(madx)
[[Variable] L = 5, [Command] QF: QUADRUPOLE {'k1': 'pi/5', 'l': 'L'}]
>>> Parser.parse(madx)
[[Command] QF: QUADRUPOLE {'k1': 0.6283185307179586, 'l': 5}]


The MADX Engine API can be used to run MADX scripts. The MADXEngine class expects a set of templates which will be used to run the script. A template is a MADX script that contains unfilled parts which can be interpolated later on. The first template is considered the entry point (the main script) and will be run.

The following code creates an engine:

from madplot.madx.engine import MADXEngine

engine = MADXEngine(
    ['test.madx', 'test.seq'],  # Template files; `test.madx` is the main script.
    madx='/opt/madx',  # File path to the MADX executable; if not specifed the `MADX` environment variable will be considered.
    working_directory='/tmp/test'  # The directory in which the engine runs the scripts.

The templates can contain substitutions following the Python string formatting rules. For example: QF: QUADRUPOLE, KL={kl};. The {kl} part can be interpolated when running the scripts.

The run method can be invoked to run a script. It expects a list of output file names (which need to be generated by the template scripts). By default the file contents will be returned as pandas.DataFrame instances.

twiss, =['example.twiss'])

Here the file example.twiss needs to be generated when running test.madx. In case one or more template scripts require interpolation the corresponding values can be specified using the configuration keyword argument:

twiss, =
    configuration={'test.madx': {'kl': 0.01}}

Special arguments for the output conversion can be specified per output in form of a dict:

(twiss, meta), =[('example.twiss', {'return_meta': True}])

This will return meta data (prefixed with @ in the TFS output) along the main data frame.

Running without creating intermediary files

The MADXPipe class runs scripts without creating intermediary script files. This is useful in order to minimize the load on the file system. It yields stdout and stderr from the underlying MADX sub-process:

from madplot.madx import MADXPipe

runner = MADXPipe(madx='path/to/madx')
with open('example.madx') as fh:
    stdout, stderr =

Templating and formatting is done manually in Python before providing the full script to the runner instance:

with open('template.madx') as fh:
    stdout, stderr = % {'h1_kick': 0.001})


The MADXSession can be used to run interactive MADX sessions. This is advantageous to avoid rerunning parts of a script that are the same for each run (e.g. sequence structure); also it doesn’t require starting a new process for each run. Instead one can only issue the relevant commands (e.g. update an optics parameter) and then ask for the results (e.g. Twiss file generation). For example:

from madplot.madx.engine import MADXSession

with open('/tmp/log', 'w') as log:
    session = MADXSession(stderr=log, stdout=log)['a := ranf()'])['value a'] * 3)

# Running a script at start-up.
session = MADXSession(['twiss_script.madx'])
twiss, =['example.twiss'])
# Update a parameter and regenerate twiss.
twiss, =['some_parameter = 0', 'twiss, file="example.twiss"'],

Using Jinja as templating engine

The JinjaEngine and JinjaPipe classes allow for using the Jinja2 templating engine for configuring single runs. JinjaEngine creates intermediary script files for each configuration, similar to the MADXEngine class, while JinjaPipe directly pipes input and output to the MADX sub-process, similar to MADXPipe.

from random import random
from madplot.madx import JinjaEngine, JinjaPipe

file_runner = JinjaEngine('example.madx.j2', madx='path/to/madx')
twiss, =['twiss.tfs'],
                         configuration={'quadrupole_gradient_errors': {f'quad_{i+1}': 0.001 * random() for i in range(18)}},

pipe_runner = JinjaPipe('example.madx.j2', madx='path/to/madx')
stdout, stderr ={f'quad_{i+1}': 0.001 * random() for i in range(18)})


Various functions for plotting are available in the madplot.plot module. Please refer directly to this module for further information.


Utilities for conversion of data formats are available at madplot.utils:

  • Convert.tfs: Converts TFS file to pandas data frame,
  • Convert.trackone: Converts trackone table (as outputted by TRACK, onetable = true) to pandas data frame.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for madplot, version 0.4.1
Filename, size File type Python version Upload date Hashes
Filename, size madplot-0.4.1-py3-none-any.whl (19.4 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size madplot-0.4.1.tar.gz (20.4 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page