ambiguity - pannous/angle GitHub Wiki
Mantra
How is the battery included approach compatible with the good old Python mantra: There should be one -- and preferably only one -- obvious way to do it.
significant whitespace
Not directly an ambiguity but something to be aware of:
⚠️ Angle has a few situations where a white space changes the semantics
resolution: interactive
Ambiguity should ideally be resolved by the compiler by asking questions.
a {x=1} Did you mean a:{x=1} or [a,{x=1}] ?
a>b:c Did you mean a>b : c or a > b:c ?
significant whitespace
significant-whitespace / semantic resolution
single element lists
in general in wasp lists plus element means extending the list:
(x,y)+z=(x,y,z)
in general in wasp lists of one element are identical to the element:
(x)=x
however this might lead to conflicts, eg in list extensions:
(x)+y=(x y) ≠ x+y
resolvable by the parser?
Different list separators have different bindings: While 1,2,3 binds as one expression, 1;2;3 yields the last element as return value.
grouping vs application
Consider a b (a) (b) {a} (b) and {a}(b).
Are they the same? A group with one element is identical to the element itself:
Axiom (a) == a
Is a block with one element identical to the element itself?
{a} == a ?
Is {a} (b) a list with elements {a} and b or is it b applied to {a} {a}(b) ?
grouping
while the precedence of ',' and '; is clear:
1,2;3,4 == ((1,2);(3,4)) the precedence of ',' and ' ' less is clear: 1 2 3 , 4 5 6and1,2,3 4,5,6`
both seam like reasonably natural groupings.
Problematic is the ambiguity in mixing 'and', comma and other operators
x, y == a, b
x, (y == a) , b
The last reading is sometimes the desired one, e.g. in argument lists
x, y as point == x, (y as point) or x, y as point == (x, y) as point == point(x,y)
[Todo](/pannous/angle/wiki/Todo)! What to do?
For colons the situation is dire as well:
`a b c:d` versus
`to skin a cat: do something`
[significant-white](/pannous/angle/wiki/significant-white) space to the rescue!
### burden of calls withoug parens
`square 3` yields a certain [beauty](/pannous/angle/wiki/beauty)
However when mixing function calls and operators, things can get ambiguous quickly:
`square 3 * 3 ` can be read as `square (3 * 3)` or `(square 3) * 3`
`square 3 + square 3` can be read as `square (3 + square 3)` or `(square 3) + square 3`
Operators [bind](precedence) more tightly than functions.
Only the last case where functions are inter-mixed needs special attention:
### resolution: braces via IDE
One way to resolve ambiguities is by making the IDE suggest and insert braces
square 3 + 4 times 3 square(3 + (4 times 3))
Even though there might not be an ambiguity to the compiler, it can be helpful to insert braces to make it easier for the user to parse the meaning:
`a or b and c => a or (b and c)`
The reasoning behind this is that the developer might not always we be aware of the [function](/pannous/angle/wiki/function) [operator](/pannous/angle/wiki/operator) [precedence](/pannous/angle/wiki/precedence).
### ambiguity in c
```while (char c = dest[i] and i>0) ```
sets c to 0 or 1 because and binds stronger than assignment!!
## ambiguous referenceIndices
### ambiguous it keyword
In [lambda](closures) expression, the `it` keyword refers to the first (implicit argument), otherwise it refers
to [self](/pannous/angle/wiki/self) or the last [result](/pannous/angle/wiki/result), which can be very ambivalent:
double=it2 # ok, interpret immediately? double:it2 # may interpret to anything later double:=it*2 # ok, implicit argument for function 'double'
### functions or functors
```map [1 2 3] square``` could be read as
```map [1 2 3] :square``` symbolic function pointer
```map ([1 2 3] square) == map [1 4 9]```
## variables vs functions
ambiguous [assignment](/pannous/angle/wiki/assignment) concepts:
name, age, source = "Crystal, 123, GitHub".split(", ") # deconstruction assignment name age, source := "Crystal, 123, GitHub".split(", ") # function 'name' with arguments age, source
The last case is especially problematic with
# Multi word variable names and typed variables
name age := age+1 The function 'name' with parameter age could also be read as name_age or age(type=name)
### block evaluation
Ideally we could reconcile `f={a:1 b:2} f[a] == 1` with the current mechanism `f(a=3) == {a:3 b:2}` or `f(a=3) == b:2`
or `f(a=3)==2` because the last expression of the block is returned.
[Map](/pannous/angle/wiki/Map) [match](/pannous/angle/wiki/match)ing versus [block](/pannous/angle/wiki/block) [selection](/pannous/angle/wiki/selection) versus [construction](/pannous/angle/wiki/construction).
### is not
`a is not b // ambiguity: a == !b vs a != b`
### Broadcasting
Broadcasting applies to functions but not to basic operators:
[1,2,3]+4 == [1,2,3,4] [1,2,3]>>1 == [2, 3] [1,2,3]++ == [2, 3]
operators √ ^
Because √ and ^ lack graphical grouping or closure in text, their usage is ambivalent:
√1+2 ≠ √(1+2)
2^x+1 ≠ 2^(x+1)
The default is that they are strongly binding (high precedence), but a warning shall be emitted!
similar operators
3 ≈ 3.1
In an early talk Ken was explaining the advantages of tolerant comparison. A member of the audience asked incredulously, “Surely you don’t mean that when A=B and B=C, A may not equal C?” Without skipping a beat, Ken replied, “Any carpenter knows that!” and went on to the next question.
implicit multiplication
2x == 2*x OK only without space: [2 x] = [2,x] BUT 2 km??
modifying functions
Can wasp solve the problem of recognizing mutating functions vs pure functions? The problem: it is a priority not known whether a method modifies the object or returns and new object.
"abc".upper()
[1,2,3].append(4)
There are some soft guidelines: • if a method returns something it's unlikely to also modify the object and conversely • if a method does not return anything it's highly likely that the object was modified.
However there are exceptions to the rule (like pop() in python) and it remains a constant source of confusion and errors.
// todo 'do' notation to modify versus return different list!
assert_emit("pixel=[1 2 3];do add 4 to pixel; pixel", Node(1, 2, 3, 4, 0));