Extending the Python ecosystem

ideas.md

Extending the Python ecosystem

1. SQLAlchemy.

Unfortunately, ArangoDB isn't SQL, so this would probably have to expose AQL instead of SQL, or only support a subset. The question is whether SQLAlchemy's architecture is flexible enough for AQL instead of SQL.

2. Ibis

It provides a dataframe API that compiles to backend-specific query languages.

https://ibis-project.org/

t = con.table("users")

t.filter(t.age > 30).group_by("country").count()

3. Pandas connector

df = arangodb.to_pandas()

4. Polars connector

db.aql(...).to_polars()

5. Apache arrow

Arrow has become the common interchange format.

6. NetworkX

I could take over the project. Update the CI, at first.

https://github.com/arangodb/nx-arangodb

It has 3 issues and 2 draft PRs

I could also add better support in the driver:

G = db.graph("social").to_networkx()

7. Pydantic models

Very Pythonic.

class User(BaseModel):
    name: str
    age: int

db.collection("users").insert(User(...))

users = db.collection("users").find(User)

This is becoming the standard experience in modern Python libraries.

Note: for these extensions, use optional dependencies in pyproject.toml:

[project.optional-dependencies]
pandas = [
    "pandas>=2.0",
]