Clone and Run: What requirements.txt Is For

Clone and Run: What requirements.txt Is For#

A large share of this course comes down to one promise: if you clone my repository and follow a few commands, you will get the results I got. This page walks through those commands with the HW 0 repository.

The four steps#

git clone https://github.com/finm-32800/hw0.git
cd hw0
conda create -n finm python=3.12
conda activate finm
pip install -r requirements.txt

  1. git clone copies the repository, including its entire history, to your computer.

  2. conda create makes a new, empty virtual environment named finm with its own copy of Python 3.12. Nothing you install into it can disturb the other Python projects on your machine.

  3. conda activate switches your terminal into that environment.

  4. pip install -r requirements.txt installs the packages that the project lists in its requirements.txt file.

Now get the data and run the project’s dashboard:

doit
streamlit run src/app.py

doit is a task runner. It reads the file dodo.py, which lists the steps of the pipeline and what each one depends on: get the data, then execute the notebooks. We study it properly in week 2.

A browser window opens with an interactive version of this week’s analysis. You did not write any of that code, and you did not have to guess which packages it needs. The repository told your computer what to install.

Why the version numbers matter#

Open requirements.txt. Each line names a package and an exact version:

numpy==2.5.3
pandas==2.2.3
streamlit==1.64.0

The == is a pin. Compare it with a floor, such as pandas>=2.0, which accepts any newer version. A floor is convenient. A pin is reproducible. Software changes, and sometimes a new version of a package changes your numbers without raising an error. One example from the packages used in finance classes: the yfinance package changed the default of the auto_adjust option of its download function from False to True. After the upgrade, the same line of code returned prices adjusted for dividends and splits where it used to return raw closing prices, and any returns computed from that column changed with it.

In this week’s paper, small differences matter. The optimal portfolio weights are very sensitive to their inputs. If your collaborator’s environment differs from yours, you may both be “right” and still disagree.

Discussion

Look closely at the pinned versions. The wrds package requires an older version of pandas than the newest one available. pip worked that out for you when it resolved the dependencies. What would happen to this project in two years if the versions were not pinned?

Different projects, different environments#

Try a second repository:

git clone https://github.com/jmbejara/streamlit_finance_chart.git

It has its own requirements.txt, with different packages and versions. Installing both projects into the same environment will eventually produce a conflict. The habit to build is one environment per project. We cover this properly, including conda environment files and the newer tools uv and pixi, in Virtual Environments in week 1.

Contents