This is a brief summary of the progress made on PyVRP in 2024, and a look at some things planned for 2025. Read more (2 min)

I figured it would be good to commit some of my thoughts, wishes, and hopes for the new year to paper, just like I did last year. Read more (1 min)

I figured it would be good to commit some of my thoughts, wishes, and hopes for the new year to paper. Doing that here lets me look back in twelve months to see how I did, so here goes nothing! Read more (1 min)

Let $X, Y \subseteq \lbrace 0, 1, \ldots, N \rbrace$. These sets can be written in equivalent form as bit arrays of size $N + 1$. Let $x$ be the bit array form of $X$ (and similarly, $y$ of $Y$): the bit $x_i$ is flipped iff $i \in X$, and not set otherwise (where $i \in \lbrace 0, \ldots, N \rbrace$). Such bit arrays enable the following set operations:

• The membership test of $i \in X$ is given by (x >> i) & 1.
• The complement $X^\complement$ is given by ~x.
• The intersection $X \cap Y$ is given by x & y.
• The union $X \cup Y$ is given by x | y.
• The relative complement $X \setminus Y$ is given by x & ~y.
• The symmetric difference $X~\Delta~Y$ is given by (x & ~y) | (y & ~x), which is equivalent to x ^ y.

These operations can be performed very efficiently in hardware, which sometimes offers substantial performance gains over other approaches (e.g., those based on hash sets).

First post using Nikola. I use this post to compare how things render in various themes. Read more (1 min)