Midas Python API

The midas Python package is in $MIDASSYS/python. Most of it wraps the midas C++ library, so you get pythonic access to an experiment without having to deal with handles, buffer sizes or status codes: functions return normal Python values (or numpy arrays), and errors are raised as exceptions. The file reader is pure Python, and works without the rest of midas.

Modules

Most programs only need midas.client (or midas.frontend for a frontend).

Creating a client

Always create your MidasClient (or your frontend object) with a with statement:

import midas.client

with midas.client.MidasClient("my_program") as client:
    state = client.odb_get("/Runinfo/State")

This guarantees that your program disconnects cleanly from midas when the with block ends. It also guarantees that any cleanup functions you registered are called, even if there's an uncaught exception or your code calls exit(). Without the with statement, Python may not disconnect your program cleanly. See Doing work when the program exits for details.

Requirements

• midas.client, midas.frontend and midas.sequencer need Python 3. They use the midas C++ library, so midas must have been compiled with CMake.
• midas.file_reader and midas.event are pure Python. They can read midas files on a machine without a midas installation.

The client and frontend libraries are designed to be easy to use, rather than as fast as possible. If performance is critical, write the program in C++, or use numpy arrays (see below).

Supported features

The Python modules cover almost everything that midas programs need. Only a few features of the C++ API are missing, e.g. writing elog entries. midas.client supports:

• ODB access: creating, deleting and renaming keys, getting, setting and watching values, and creating links
• run transitions: starting, stopping, pausing and resuming runs, and registering callback functions for each transition
• event buffers: connecting, sending events and receiving events
• remote procedure calls: connecting and talking to other midas clients
• alarms: creating and triggering alarms
• messages: sending messages and getting recent messages
• history: reading data

Using numpy

The Python libraries can use numpy arrays for ODB values and for the data in midas events (from event buffers or midas files). If the array has the numpy type that matches the midas type, the data is passed to the C++ library without any conversion, which is much faster.

  def __init__(self):
-   self.fake_data = [0] * 1250000
+   self.fake_data = np.ndarray(1250000, np.int16)

  def readout_func(self):
    event = midas.event.Event()
    event.create_bank("MYBK", midas.TID_SHORT, self.fake_data)
    return event

The dict midas.tid_np_formats gives the numpy type for each midas type. For example midas.tid_np_formats[midas.TID_SHORT] is np.int16. If you write an int16 array to a TID_SHORT bank, the fast path is used. If the types don't match, the data has to be converted, which is much slower.