anubh-v / non-deterministic-source Goto Github PK

Currently, the REPL has no memory, i.e names declared in an iteration of the loop are only accessible in the iteration in which they are declared. In subsequent iterations, an Unbound name error occurs when the name is accessed.

Having the ability to reference names declared in previous iterations will increase the ease with which more complex programs can be written.

To reproduce:
parse_and_eval("const f = 1; f = 3; f;");

Expected: An error message indicating we cannot mutate the value of f

Observed: Value of f becomes 3

This is not a bug, rather a limitation.
This mechanism did not exist in the evaluator that we started from

Currently, we lack support for:

• Arrays (array creation via literal array syntax, array access and array assignment)
• for / while loops
• break and continue keywords

Need to decide if this in within scope

This is required for baseline goals.
Do we include as many as we can / all programs in the textbook?

Here are some of the simpler programs:

• is_element_of

• an_integer_starting_from (present in existing tests)

• an_integer_between

• prime_sum_pair

• a_pythagorean_triple_between

• Logic puzzle on multiple dwellings (requires the distinct list function)

Currently, a global variable final_result is used to capture the result of an amb statement. When try_again is invoked, final_result is updated with the new value while try_again returns undefined.

Having try_again return a value can provide a cleaner way of accessing the new value.

Currently, testing of functionality that involves errors is not feasible. This is due to the error function which aborts the program.
As such, instead of invoking error when needed, the following can be done:

Keep track of the error message in a global data structure.
Subsequently check the error messages stored in the data structure as part of the tests in order to verify the expected behaviour of the program.

Eg.

const f = x => x + 10; f(12);

Error shown: Line 534: Expected 3 arguments, but got 1.

The bug may lie in analyze_return