Skip to main content

Programming Languages

Programming-based approaches are the 'gold standard' for reproducibility, as every step of your analysis is documented in code.

R​

R is a language specifically designed for statistical computing and graphics.

  • Best for: Complex statistical modelling and publication-quality visualisations.
  • TUM Context: Widely used in Statistics, Bioinformatics, Life Sciences and Social Sciences.
  • Data Visualisation with ggplot2:
    • The ggplot2 package is the industry standard for R visualisations.
    • It is based on the 'Grammar of Graphics', which allows you to build plots layer by layer (data + aesthetics + geometric objects).
  • Core Resource: We highly recommend Gagneur Lab: Data Analysis and Visualisation in R.

Python​

Python is a general-purpose language that has become the dominant tool for Machine Learning, Artificial Intelligence, and Data Engineering.

  • Package Management with pip:
    • pip stands for 'Pip Installs Packages' (a recursive acronym). It is the standard package manager for Python.
    • uv has established itself as the superior and vastly faster alternative to pip as a package installer, venv as a virtual environment manager and other tools such as pyenv or poetry as dependency managers
    • Best Practice: Always use Virtual Environments (uv, venv or conda) to keep your research project dependencies isolated.
  • Dependency management:
    • most code is dependent on other libraries / Python packages
    • as libraries change, code that depends on these libraries might become incorrect over time. It is therefore essential to not only know on which libraries your code depends on, but also what version of these dependencies it assumes.
    • the Python Foundation has developed an open standard to denote dependencies in a file called pyproject.toml. It contains (alongside other metadata) a list of dependencies and their (usually minimal) version requirements
    • pyproject.toml does not need to be handwritten. Tools such as uv manage it for you.
    • a lock file is a file that contains the exact version numbers and checksums of all currently installed dependencies including all transient dependencies. Such a file is necessary in order to exactly reproduce a Python environment and should definitely be included in scientific analysis code. The lock file uv produces is called uv.lock.
  • Essential Libraries: Pandas (data), Polars (data, viz), Scikit-learn (ML), PyTorch (AI), and Matplotlib (viz).

C and C++​

These are compiled languages used when performance is the highest priority.

  • Best for: Developing high-performance software, simulations, and compute-intensive algorithms that need to run on HPC clusters.
  • TUM Context: Core to Engineering, Physics, and the development of new bioinformatics tools (like sequence aligners).
  • Note: While harder to learn than Python, they offer significantly faster execution speeds for massive datasets. However, slower languages can still benefit from the speed of compiled C/C++ code via so-called bindings. Bindings allow you to make calls to compiled code from other languages such as Python or R. This is how Python libraries like numpy achieve their speed.

JavaScript​

JavaScript is the language of the web, increasingly used in research for sharing results.

  • Best for: Creating interactive web-based visualisations and data dashboards.
  • Key Libraries: D3.js (complex data-driven documents) and React (building user interfaces).
  • Usage: Essential if you are building a web tool for other researchers to explore your findings.

SQL​

Structured Query Language (SQL) is the standard for managing and querying data held in relational database management systems.

  • Best for: Interacting with large enterprise data warehouses and structured datasets.