Repository navigation
Arithmetic that differs from Python: % with negatives, ** grouping, round halves, bool as int, e #224
Description
While fixing #170 I ran a 122-expression arithmetic program through Web VPython 3.2 and compared every result with CPython. 107 match. The rest differ. All of these differences are older than #170: the release before it has every one of them.
| Expression | Web VPython | Python | Cause |
|---|---|---|---|
7 % -3 |
1 | -2 | % is JavaScript's remainder, which takes the sign of the left operand. Python's takes the sign of the right one. |
-7 % 3 |
-1 | 2 | same |
-7.5 % 2 |
-1.5 | 0.5 | same |
a = 7; a %= -3 |
1 | -2 | same |
2**3**2 |
64 | 512 | ** groups left to right (RapydScript emits Math.pow(Math.pow(2, 3), 2)). Python groups it right to left. |
2**-1**2 |
0.25 | 0.5 | same |
round(2.5), round(-2.5) |
3, -3 | 2, -2 | Halves round away from zero. Python rounds them to the even neighbour. |
int(True), True + 1 |
error | 1, 2 | Booleans aren't treated as the integers 0 and 1. |
log(e) |
error | 1 | e isn't defined (pi is). |
a **= 2 |
compile error | "Unexpected token: operator «=»" |
The division-by-a-negative cases from #170, true and floor division, chained comparisons, abs/min/max/sum/pow/sqrt/divmod, range with negative steps, the trig and log functions, and integers up to 2**52 all match Python.
I also found a regression introduced by #170. Fixed with a test in #225. Before #170, the "-" → "-1*" rewrite hid a RapydScript bug: it printed - -x as --x, which in JavaScript is a decrement (wrong answer, and x changes). - -3 didn't compile at all.
Related: #217 (refreshing the embedded RapydScript). I haven't checked which of these upstream RapydScript-NG has fixed; the ** grouping is still the same at upstream 3e69ba8.
The Trinket deploy suite pins these as known differences (PICUP-Physics/trinket-oss#369, vpython-arithmetic.spec.js), so a fix here will show up there.
🤖 Generated with Claude Code
Activity
test: check 112 Web VPython expressions against CPython
sspickle commentedon Nov 3, 2026
Update: the parser fix in #225 also corrects the
**grouping items here.2**3**2now gives 512 and2**-1**2gives 0.5. That leaves%with negatives, round halves,int(True)/True + 1,e, anda **= 2.While testing #225 I found a second #223 regression, now fixed there:
-(x+1)**2came out positive. glowscript.org is rolled back to the version before #223 until #225 is merged.🤖 Generated with Claude Code
fix: parse ** tighter than a unary operator on its left, as Python does