algolearn provides a very simple "learner computer" in Python to solve foundational algorithmic challenges that has historically been taught in Pascal, C or even Basic.

You write programs in Python and have access to a "VM" with 16 bit memory (64k in 64 byte pages). There are only limited instructions available and you need to learn about pointers.

Currently it's all quite simple, you can simply use Python 3.5 from your system without dependencies.

$ python3 l1.py <yourprogram>

There are a number of boilerplate files that you can use to get started:

$ python3 l1.py linkedlist.py

Here's the list of available instructions:

• malloc(size) -> address
• free(address)
• write(address, value)
• read(address) -> value
• write_addr(address, value_address)
• read_addr(address)

And the data types:

• values - are single bytes (8-bit, 0..255)
• addresses - are words (16-bit, 0..2^16-1)

Additionally you can access a constant:

• PAGESIZE (could vary, default is 64 bytes)

• The zero page is pre-allocated and used for system purposes (really, real machines would have those for ROMs but I just want to not use it to avoid null pointer confusion and I don't want to necessarily have virtual memory (yet))

• The VM performs a certain amount of accounting and assigns a cost to every operation and reports on it at the end. The specific cost is assigned in a "strategic" way for teaching, not for ultra realistic simulation.

• I'm pondering to add RestrictedPython to the mix so that specific VMs can limit what the students have access to. However, I do want them to be able to take some convenience shortcuts (accessing print and output formatting) so they can focus on the algorithm. I likely want to reduce access to higher order data types (lists, sets) etc so that I don't have to manually review the code for unintended overreach.

We're using this for our apprentices and if this takes off then I might create multiple VM implementations with different limitations and abilities.

I'm happy to get feedback, questions, and ideas. Simply make a Github issue.