It has been a while since I worked through it, but Haskell programming from first principles had some pretty good exercises IIRC.

You could do worse than The Art of Computer Programming.

Solve real life problems from work. It's easy to solve problems, so practice being a problem spotter.

Some especially good ones

Sipser -- Theory of computation

Pinter -- Abstract algebra

Kleppner and Kolenkow -- Mechanics

Klein -- Organic Chemistry as a Second Language

There are several principles for reference: 1. Exercise quality: Exercises must cover core concepts, have a reasonable difficulty level, and preferably come with answers or explanations. 2. Practice orientation: Deepen theoretical understanding and cultivate application capabilities by solving problems. 3. Classics and reputation: Give priority to books that have been recognized by the academic community or learners for a long time.