herepath

A simpler way to find your files. A Python port of the R here package.

📖 Full documentation (English/French): https://pinto-katende-jonathan.github.io/herepath/

herepath builds paths relative to your project's root, no matter what the current working directory is. Stop writing brittle ../../data/x.csv paths or relying on where a script happens to be launched from.

from herepath import here

here("data", "penguins.csv")
# -> /home/me/myproject/data/penguins.csv   (always, from anywhere)

Installation

pip install herepath

One name everywhere: pip install herepath, import herepath, and a herepath command. (An unrelated pyhere package by another author exists on PyPI; herepath is deliberately named to avoid any confusion with it.)

For local development, from the project directory:

pip install -e ".[dev]"

How it works

On the first call to here(), herepath walks up the directory tree from the current working directory until it finds a directory matching one of these criteria (in order):

The closest matching ancestor wins, so a sub-package in a monorepo resolves to its own directory. If nothing matches, the current working directory is used as a fallback. requirements.txt is deliberately not a marker: it is too often duplicated in subdirectories (docs/, tests/), which would anchor the root in the wrong place. Override the criteria with set_criteria() if you need to.

API

here(*args) -> Path

Drop-in replacement for os.path.join, anchored at the project root.

here()                       # the project root
here("data", "raw", "x.csv") # root/data/raw/x.csv
here("data/raw/x.csv")       # same: components may contain "/"

data = here("data")          # absolute
here(data)                   # returned unchanged
here(data, "x.csv")          # absolute anchor kept -> data/x.csv

The result is always an absolute Path, so it stays valid even if the working directory changes later.

i_am(path, *, uuid=None) -> Path

Declare where the current script lives, relative to the project root. This is the recommended and most robust way to fix the root. Put it near the top of a script or notebook:

from herepath import i_am, here
i_am("analysis/report.py")

here("data", "penguins.csv")

It walks up from the working directory until it finds a directory containing path, and pins that as the root. On success it prints a one-line report (here() starts at ...); pass quiet=True to suppress it. If the file can't be found, it raises FileNotFoundError with a descriptive message, which protects you from running a script in the wrong place. Pass uuid="..." to also require a unique marker string within the first 100 lines of the file, for extra safety.

set_here(path=".", verbose=True) -> Path

Create an empty .here marker file to pin a root when no other criterion applies.

reset()

Forget the cached root so it is re-detected on the next call. Handy in long-lived sessions like Jupyter notebooks (after moving files or changing directory) and in tests.

find_root(*criteria, start=".") -> Path

Lower-level escape hatch for custom root markers, in the spirit of R's rprojroot::find_root. Does not cache anything or touch the session root. Build criteria with has_file, has_dir, has_glob; a directory matches if it satisfies any of them:

from herepath import find_root, has_file, has_dir

find_root(has_file("Makefile"), has_dir(".git"))

set_criteria(*criteria) / reset_criteria()

Customise what counts as a project root for the whole session. Useful for organisations with their own markers:

from herepath import set_criteria, has_file, has_dir

set_criteria(has_file("company_project.json"), has_dir("src"))
# ... here() now uses these markers
reset_criteria()  # back to the built-in defaults

using_root(path) context manager

Temporarily pin the root, restoring the previous state on exit. Ideal in tests:

from herepath import using_root, here

with using_root(tmp_path):
    assert here("data") == tmp_path / "data"
# previous root (or auto-detection) restored here

Debugging detection: dr_here(trace=True)

When the wrong root is picked, trace=True prints the full upward search:

Searching from:
  /project/notebooks
Checking:
  /project/notebooks
  /project   <- contains a file `pyproject.toml`
Matched:
  /project

Forcing the root with HEREPATH_ROOT

Set the HEREPATH_ROOT environment variable to an existing directory to override auto-detection entirely. This is the recommended escape hatch for Docker, CI, and deployment, where the heuristics may not apply:

HEREPATH_ROOT=/app python analysis/report.py

If it points to a path that isn't a directory, here() raises ValueError so misconfiguration fails loudly. An explicit i_am() call still takes precedence.

dr_here(show_reason=True)

Print a situation report explaining where the root is and why, which is useful when here() gives unexpected results.

here() starts at /home/me/myproject
- This directory contains a file `pyproject.toml`
- Initial working directory: /home/me/myproject/analysis
- Current working directory: /home/me/myproject/analysis

When not to use herepath

herepath is for scripts, notebooks and analyses: code run from within a project tree. It is not meant for use inside an installed library, because once a package is installed the source layout no longer exists. To access data bundled with an installed package, use importlib.resources instead.

Differences from the R package

• Returns pathlib.Path objects instead of strings.
• The default root criteria are Python-flavoured (e.g. pyproject.toml) rather than R-flavoured (e.g. DESCRIPTION), while keeping the universal markers (.here, .git, editors).
• The root is resolved lazily on first use (Python has no package-load hook tied to a session the way R does).

Command line

Installing the package also provides a herepath command, handy in shell scripts and Makefiles:

herepath                      # print the project root
herepath data penguins.csv    # print root/data/penguins.csv
herepath --report             # situation report (why this root?)
herepath --version

# e.g. anchor a command at the project root from any subdirectory:
cat "$(herepath data/penguins.csv)"

Development

pip install -e ".[dev]"
pre-commit install   # optional: run linters on commit

pytest               # tests
pytest --cov         # tests + coverage
ruff check .         # lint
ruff format .        # format
mypy                 # type-check

See CONTRIBUTING.md for the full guide. The project is tested on CPython 3.8 to 3.13 across Linux, macOS and Windows.

Authors and acknowledgements

Main developer and creator: Jonathan Katende Pinto.

Main contributors:

• MWANZA LUBUKAYI Henock
• MUTONJI BUKAMA Arsène
• KHANG MATE ZULBAL Emmanuel
• MUKWIYO MUKALO Patrick

Inspired by the R here package by Kirill Müller and Jennifer Bryan.

License

MIT, © Jonathan Katende Pinto.