View source for Using Python from Analytica/Setting up a Python environment

This page explains how to prepare a Conda-based Python environment and configure Analytica to use it. Python 3.8+ is '''recommended''' for better compatibility and support. ''Other Python installations can be used if you provide the full Python home directory path in Analytica’s Python system variables.''

== Overview of Conda and distributions ==
Conda is a cross-platform package and environment manager that creates isolated Python environments and installs packages with precise dependency control.

=== Anaconda and Miniconda ===
[https://www.anaconda.com/download/success Anaconda and Miniconda] are distributions that include Conda.

'''Anaconda''' bundles many scientific packages and tools out of the box and includes the '''Anaconda Navigator GUI'''.

'''Miniconda''' is a minimal installer that provides only Conda and Python so that you add exactly the packages you need.

=== Selecting a Conda distribution ===
Choose '''Anaconda''' for convenience when a GUI is preferred and many packages should be available immediately; the trade-off is a heavier download and larger disk footprint.

Choose '''Miniconda''' for a lightweight installation that starts small and grows only as needed using <code>conda install</code>.

''The example below use the Conda command line which are included in both '''Anaconda''' and '''Miniconda'''.''

== Installing Conda ==
Download and run the Anaconda or Miniconda installer.

[[File:AnacondaInstaller.png]]

== Creating and managing an environment ==
Open the '''Anaconda Prompt''' and create an environment named <code>MyEnv</code> with commonly used libraries:

<code>conda create -n MyEnv python=3.11 pandas numpy</code>

Python 3.8+ is recommended; selecting a newer version such as <code>python=3.12</code> is acceptable.

=== Package sources and managers ===
About conda-forge: <code>conda-forge</code> is a community-maintained collection of Conda packages. Install from it when required, for example:

<code>conda install -n MyEnv -c conda-forge matplotlib</code>

Mixing channels can sometimes introduce dependency conflicts or larger downloads, so prefer a consistent channel strategy.

Installing some packages with <code>pip</code> inside the environment can be acceptable, but prefer <code>conda install</code> when possible for better dependency management.

== Activating the environment and installing more packages ==
Analytica does not require shell activation to use the environment, but activation is useful when installing or testing packages interactively.

To work inside the existing environment and add packages:

<code>conda activate MyEnv</code>

<code>conda install requests</code>

<code>pip install some_package_only_on_pip</code>

Packages can be added or updated at any time after the environment is created.

== Configuring Analytica to use the environment ==
Open the '''Definition / Application integration / Python''' submenu. Set either '''Python Model Environment''' or '''Python Environment'''. '''''Python Model Environment''' is saved with the model file, while '''Python Environment''' is saved in Analytica.'' Enter the environment name (e.g., <code>MyEnv</code>) or the full Python home folder path (e.g., <code>C:\Users\username\anaconda3\envs\MyEnv</code>).

When only a name is provided, Analytica searches the default Anaconda and Miniconda environment locations. If none of the Python system variables are set, Analytica looks first in the Windows User registry and then the Machine registry for a Python home. When both settings are present, '''Python Model Environment''' takes precedence over '''Python Environment'''.

''If Python was already initialized in this session, close and reopen the model or restart Analytica after changing the environment.''

== Verifying from within Analytica ==
After setting '''Python Model Environment''', run a minimal verification that reports the NumPy version:
<pre>import numpy as np
np.__version__</pre>

== Troubleshooting ==
If the specified environment name or path is not found, an [[Error_Messages/46064|error]] is raised.

For package availability or dependency issues, try installing with <code>conda install</code> first; if a package is not available on the default channels, try using an [https://docs.conda.io/projects/conda/en/latest/user-guide/concepts/channels.html alternative channel] like [https://conda-forge.org/ conda-forge] while keeping channel usage consistent. As a last resort, use <code>pip install</code> inside the environment, recognizing that this mixes package managers.