.. _algebra_performance:

Performance considerations in algebra mode
===============================================

Creating lots of Shapes in a loop means for every step ``fuse`` and ``clean`` will be called. 
In an example like the below, both functions get slower and slower the more objects are 
already fused. Overall it takes on an M1 Mac 4.76 sec.

.. code-block:: build123d

    diam = 80
    holes = Sketch()
    r = Rectangle(2, 2)
    for loc in GridLocations(4, 4, 20, 20):
        if loc.position.X**2 + loc.position.Y**2 < (diam / 2 - 1.8) ** 2:
            holes += loc * r

    c = Circle(diam / 2) - holes


One way to avoid it is to use lazy evaluation for the algebra operations. Just collect all objects and 
then call ``fuse`` (``+``) once with all objects and ``clean`` once. Overall it takes 0.19 sec.

.. code-block:: build123d

    r = Rectangle(2, 2)
    holes = [
        loc * r
        for loc in GridLocations(4, 4, 20, 20).locations
        if loc.position.X**2 + loc.position.Y**2 < (diam / 2 - 1.8) ** 2
    ]

    c = Circle(diam / 2) - holes

Another way to leverage the vectorized algebra operations is to add a list comprehension of objects to
an empty ``Part``, ``Sketch`` or ``Curve``:

.. code-block:: build123d

    polygons = Sketch() + [
        loc * RegularPolygon(radius=5, side_count=5)
        for loc in GridLocations(40, 30, 2, 2)
    ]

This again ensures one single ``fuse`` and ``clean`` call.