arrowresearch / qc0

Query Combinators to SQL compiler prototype

qc0

WARNING: EXPERIMENTAL, DO NOT USE

This project attempts to implement a Query Combinators to SQL compiler. The current emphasis is on correctness, generated SQL is not optimal in most but basic cases.

There's no concrete syntax (yet), the main user interface is a Python API (EDSL)

Trying it out

There's qc0-shell executable provided which gives an interactive Python shell with Query Combinators EDSL configured to query PostgreSQL TPC-H database.

Start interactive shell with:

% qc0-shell

Then run queries with:

>>> q.region
...   .filter(q.name == q.val('AFRICA'))
...   .select(name=q.name, nations=q.nation.name)
...   .run()

Syntax

While there's no concrete syntax yet, we still describe it here. Note that it differs slightly from the Python EDSL API (introduced in the next section) due to the nature of the embedding into Python.

Navigation queries data specified by the identifier NAME:

NAME

Composition of two queries QUERY1 and QUERY2, it performs QUERY1 first and then performs QUERY2 in its resulting context:

QUERY1 . QUERY2

Applicaton applies a query combinator to its argument queries QUERY:

COMB(QUERY...)

Syntax Sugar

The select query combinator is so common that there's selection syntax sugar available:

{ NAME: QUERY... }

This desugars into:

select(NAME: QUERY...)

Python EDSL API

Python programmers interact with qc0 by consturcting syntax trees. The main interface for this is a q object.

To produce a navigation syntax node one access an attribute of the q:

q.region

To do composition one uses >> operator:

q.region >> q.name

To apply a query combinator one calls a method on the q:

q.select(name=q.name)

Now when you construct a long query in one shot using q with >> alone can be noisy sometimes. For that reason it's also possible to produce composition using .:

q.region.name                 # q.region >> q.name
q.region.select(name=q.name)  # q.region >> q.select(name=q.name)

There's a special q.val() method which allows to build queries from Python values. Of course such queries always evaluate to constant results designated by those values passed-in:

q.val(42)
q.val("Hello")
q.val(True)
q.val({"some": ["json"]})