Utility routines

There are a number of utility routines provided by Sherpa that may be useful. Unfortunately it is not always obvious whether a routine is for use with the Object-Oriented API or the Session API.

Controlling the verbosity of Sherpa

Sherpa uses Python logging for most messages. This allows the user to redirected the output to a file or suppress it by setting the logging level. The following example will globally change the level for all sherpa modules, such that debug and informational messages are no longer displayed:

>>> import logging
>>> sherpalog = logging.getLogger('sherpa')
>>> sherpalog.setLevel('WARNING')

Sherpa also provides a context manager - SherpaVerbosity - to change the logging level only for a specific portion of the code. This can be used, e.g., to hide the long default output printed after fitting a model:

>>> from sherpa.utils.logging import SherpaVerbosity
>>> import numpy as np
>>> from sherpa.astro import ui
>>> ui.load_arrays("mydata", np.arange(5), np.ones(5))
>>> ui.set_model("mydata", "polynom1d.poly")
>>> with SherpaVerbosity('WARNING'):
...    ui.fit("mydata")

Random Numbers

Sherpa uses random numbers

The primary way to control the random numbers used by Sherpa is to create a NumPy Random Generator and pass it to routines either with the rng parameter or, for users of the Session class, the set_rng method. However, for the optimiser code, which uses a C++ implementation of the Mersenne Twister random number generator, the important value is the seed argument (although this can be derived from the random number generator if the seed value is set to None).


Prior to Sherpa 4.16.0 the random numbers were controlled by a combination of the legacy NumPy random-number API, that is, calling numpy.random.seed() , and - in a few places, the Python random module, as well as the seed argument for the optimiser code.