This is a pseudocode to Python 3 transpiler for the OMSCS course CS6515 Graduate Algorithms.
CS6515 requires you to use 1-based indexing and a heavily-constrained pseudocode to write dynamic programming problems on homework and exams. Many students practice DP problems in Python and lose points on homework and exams due to use of forbidden syntax and language features, lack of fluency with the pseudocode or 1-indexing or simply ignoring requirements.
Rene transpiles the pseudocode to Python, enforcing the correct syntax.
This language is named Rene after a frog that appeared in some homework problems when I took the course. Everyone loves Rene.
See lcs.rene
for example code and rene.lark
for the grammar.
This is a heavily constrained Python-like (whitespace significant) toy language that attempts to disallow any syntax or features that aren't allowed on homework or exams. Notably:
- Your tools are variables, numbers, fixed-size arrays, loops, conditions and function calls. You can call plain functions like
print
but not methods likemath.log2
.- It's a good idea to restrict your function calls to those in the spirit of pseudocode,
max
,min
,abs
and so forth.
- It's a good idea to restrict your function calls to those in the spirit of pseudocode,
- Instead of augmented assignments like
+=
and unary operations like++
, usefoo = foo + 1
. - Instead of dicts or lists, use fixed arrays.
- Instead of
len(iterable)
, add explicit length variables to the function header. - Instead of
for i in range(0, n):
, usefor i = 1 -> n:
, wheren
is inclusive and transpiles tofor i in range(1, n + 1)
. - Instead of
and
andor
, use&&
and||
. - 4-space indentation is required.
- Only
/
is allowed for division.//
is a comment. If you need floor division, addfrom math import floor
and callfloor()
explicitly, but you probably won't need this, or floating point anything.
Some syntactical restrictions are due to my own ignorance. Please PR if you can fix these:
- Blank lines are allowed, but the spaces in the lines need to match the current indentation level.
else if
,elif
andelsif
areelseif
in this language. Changing this by hand toelse if
is probably safer for homework submissions. You probably won't need to useelse if
.
There is no semantic analyzer in Rene, only parsing errors, so it's up to you to debug in Python, which will mostly look like the Rene code other than line numbers and a few extra array conversion calls.
Rene uses an Array
type as a wrapper for making NumPy array
s. Array
prepends an extra row of uninitialized values on every dimension, giving you the option to index at 0 and enabling 1-based indexing otherwise.
There are two function calls that make 1-indexed arrays:
Array(*dimensions)
(alias:Table
): this array is uninitialized so you'll need to write explicit loops to set values. You can useINFINITY
andNEGATIVE_INFINITY
instead offloat("inf")
.array_from_iterable(it)
: converts an iterable to a 1-indexed iterable. You won't need to call this; the transpiler will insert calls for you on anyArray
parameters. Currently, Rene doesn't generate code to stop you from illegally accessing index 0 on these parameters since it's the same structure as your arrays/tables, so take care.
Rene does support strings but they're not 1-indexed. You could call s = array_from_iterable(s)
, but strings are mainly available for debugging messages rather than DP logic. Use an Array
parameter if your function receives a string to make it 1-indexable.
Python 3, Lark and NumPy. pip install lark-parser numpy
and download or clone this repo.
python3 rene.py lcs.rene
python3 rene.py lcs.rene lcs.py
import rene
# from source file to string
py_code = rene.generate_code(source_file="lcs.rene")
# from source file to out file and string
py_code = rene.generate_code(source_file="lcs.rene", out_file="lcs.py")
# from source code string to out file and string
py_code = rene.generate_code(source_string='print("hello")\n', out_file="hello.py")
# from source code string to string
py_code = rene.generate_code(source_string='print("hello")\n')
If you want to run your code in a test harness, see lcs_test.py
and run it with python3 lcs_test.py
. It might be smart to write your code to file when you run tests so you can look at it to debug errors (yes, this is not fancy).
Not really intended, but easy enough. Leetcode has NumPy, so you can run python3 rene.py lcs.rene
and pipe to clipboard, paste into Leetcode's editor and call your solution from their stub:
class Solution:
def longestCommonSubsequence(self, text1: str, text2: str) -> int:
return lcs(len(text1), len(text2), text1, text2)
There are command-line submission tools for LC that could automate this. Efficiency might be impaired due to importing NumPy without taking advantage of vectorized operations.
Coming soon
- improve comment and multiline string support
- support unary minus?
- add tests
- support named parameters for function calls
- make into a package for easier install
No guarantees expressed or implied. Use this software entirely at your own risk.
Yes, please and thanks!