Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 23:09:14 UTC, Paul Backus wrote:
On Saturday, 14 August 2021 at 14:04:47 UTC, kdevel wrote:
~~~
char [7] d7 = "x"; // okay
string s = "x";
char [7] c7 = s; // throws RangeError
~~~
What justifies that the compiler behaves differently on two
terms ('s', '"x"') which are of equal size, type, length and
value?
Literals in D can have different types in different contexts.
For example:
```d
byte b = 16; // 16 is treated as a byte literal
int n = 16; // 16 is treated as an int literal
b = n; // Error: cannot convert int to byte
```
The wording of the error message
~~~d
void main () // bi.d
{
byte b = 257;
}
$ dmd bi.d
bi.d(3): Error: cannot implicitly convert expression `257` of
type `int` to `byte`
~~~
does not seem to support your interpretation. The term `257` does
not encode a ‘polymorphic’ entity but the int value 257 which is
used to initialize a variable. If necessary, the value is
converted. The literal is not “typeless” as in Go [1] either.
Similarly, the string literal `"x"` can be treated either as a
`string` (a dynamic array of `immutable(char)`) or as a static
array of `char`, depending on the type of variable it's
assigned to.
[1] https://blog.golang.org/constants
Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 14:04:47 UTC, kdevel wrote:
~~~
char [7] d7 = "x"; // okay
string s = "x";
char [7] c7 = s; // throws RangeError
~~~
What justifies that the compiler behaves differently on two
terms ('s', '"x"') which are of equal size, type, length and
value?
Literals in D can have different types in different contexts. For
example:
```d
byte b = 16; // 16 is treated as a byte literal
int n = 16; // 16 is treated as an int literal
b = n; // Error: cannot convert int to byte
```
Similarly, the string literal `"x"` can be treated either as a
`string` (a dynamic array of `immutable(char)`) or as a static
array of `char`, depending on the type of variable it's assigned
to.
Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 14:04:47 UTC, kdevel wrote:
On Saturday, 14 August 2021 at 13:01:13 UTC, frame wrote:
I would say case [...] 3 is not [a bug]. It's just the element
type conversion and mismatched lengths of the ranges.
~~~
char [7] d7 = "x"; // okay
string s = "x";
char [7] c7 = s; // throws RangeError
~~~
What justifies that the compiler behaves differently on two
terms ('s', '"x"') which are of equal size, type, length and
value?
One is pure initialization, CTFE magic for strings or something
like that. The other one is an assignment of an already created
range where length check kicks in. The compiler is fine with that
as it compiles. It's the bound checker at runtime level that
bails out.
Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 14:33:42 UTC, kdevel wrote:
On Saturday, 14 August 2021 at 14:18:57 UTC, Tejas wrote:
On Saturday, 14 August 2021 at 14:04:47 UTC, kdevel wrote:
On Saturday, 14 August 2021 at 13:01:13 UTC, frame wrote:
I would say case [...] 3 is not [a bug]. It's just the
element type conversion and mismatched lengths of the ranges.
~~~
char [7] d7 = "x"; // okay
string s = "x";
char [7] c7 = s; // throws RangeError
~~~
What justifies that the compiler behaves differently on two
terms ('s', '"x"') which are of equal size, type, length and
value?
They are not.
assert(string.sizeof != char.sizeof);
The terms are s and "x":
~~~r.d
void main ()
{
string s = "x";
assert (s.sizeof == "x".sizeof);
}
~~~
It did implicit conversion on the literal but not the variable.
Basically,
```d
char[7] a = cast(char[7])"x"; // works
string s = "x";
char[7] b = cast(char[7])s; //fails
```
That's my best guess.
Hopefully someone more experienced chimes in.
Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 14:18:57 UTC, Tejas wrote:
On Saturday, 14 August 2021 at 14:04:47 UTC, kdevel wrote:
On Saturday, 14 August 2021 at 13:01:13 UTC, frame wrote:
I would say case [...] 3 is not [a bug]. It's just the
element type conversion and mismatched lengths of the ranges.
~~~
char [7] d7 = "x"; // okay
string s = "x";
char [7] c7 = s; // throws RangeError
~~~
What justifies that the compiler behaves differently on two
terms ('s', '"x"') which are of equal size, type, length and
value?
They are not.
assert(string.sizeof != char.sizeof);
The terms are s and "x":
~~~r.d
void main ()
{
string s = "x";
assert (s.sizeof == "x".sizeof);
}
~~~
Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 14:04:47 UTC, kdevel wrote:
On Saturday, 14 August 2021 at 13:01:13 UTC, frame wrote:
I would say case [...] 3 is not [a bug]. It's just the element
type conversion and mismatched lengths of the ranges.
~~~
char [7] d7 = "x"; // okay
string s = "x";
char [7] c7 = s; // throws RangeError
~~~
What justifies that the compiler behaves differently on two
terms ('s', '"x"') which are of equal size, type, length and
value?
They are not.
assert(string.sizeof != char.sizeof);
Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 13:01:13 UTC, frame wrote:
I would say case [...] 3 is not [a bug]. It's just the element
type conversion and mismatched lengths of the ranges.
~~~
char [7] d7 = "x"; // okay
string s = "x";
char [7] c7 = s; // throws RangeError
~~~
What justifies that the compiler behaves differently on two terms
('s', '"x"') which are of equal size, type, length and value?
Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 13:01:13 UTC, frame wrote: On Saturday, 14 August 2021 at 09:40:54 UTC, kdevel wrote: Shall I file them to bugzilla? I would say case 2 and 3 are not bugs. It's just the element type conversion and mismatched lengths of the ranges. Not sure about the first one. Yeah, I jumped the train :( Replace ```string``` with ```char``` and double quotes with single quotes then case 2 and 3 give the expected output 1 is 100% a bug, please report that.
Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 09:40:54 UTC, kdevel wrote: Shall I file them to bugzilla? I would say case 2 and 3 are not bugs. It's just the element type conversion and mismatched lengths of the ranges. Not sure about the first one.
Re: partial initialization of fixed size ("static") arrays
On Saturday, 14 August 2021 at 09:40:54 UTC, kdevel wrote:
Today I stumbled across three issues with partial
initialization of "static" arrays:
~~~i1.d
void main ()
{
char [7] b = [ 1: 'x' ]; // okay
char [7] a = [ 0: 'x' ]; // i1.d(4): Error: mismatched array
lengths, 7 and 1
}
~~~
~~~i2.d
import std.stdio;
void main ()
{
char [7] c7 = [ 1: 'x' ];
writeln (cast (ubyte [7]) c7); // [255, 120, 255, 255, 255,
255, 255] okay
char [7] d7 = "x";
writeln (cast (ubyte [7]) d7); // [120, 0, 0, 0, 0, 0, 0]
would have expected
// [120, 255, 255, 255, 255,
255, 255]
}
~~~
~~~i3.d
// i3.d
void main ()
{
string s = "x";
static assert (is (typeof (s) == typeof ("x")));
assert (s == "x");
char [7] c7 = s; // throws RangeError
}
$ dmd -g i3
$ ./i3
core.exception.RangeError@i3.d(7): Range violation
??:? _d_arrayboundsp [...]
i3.d:7 _Dmain [...]
~~~
Shall I file them to bugzilla?
Static arrays really don't get any love :(
I also recently filed a bug:
https://issues.dlang.org/show_bug.cgi?id=22198
I think you should also file these ones.
If they're wrong, it'll just get ```RESOLVED INVALID```. Even
better, someone here can tell why the above mentioned behaviours
aren't bugs and you can just do it yourself.
partial initialization of fixed size ("static") arrays
Today I stumbled across three issues with partial initialization
of "static" arrays:
~~~i1.d
void main ()
{
char [7] b = [ 1: 'x' ]; // okay
char [7] a = [ 0: 'x' ]; // i1.d(4): Error: mismatched array
lengths, 7 and 1
}
~~~
~~~i2.d
import std.stdio;
void main ()
{
char [7] c7 = [ 1: 'x' ];
writeln (cast (ubyte [7]) c7); // [255, 120, 255, 255, 255,
255, 255] okay
char [7] d7 = "x";
writeln (cast (ubyte [7]) d7); // [120, 0, 0, 0, 0, 0, 0]
would have expected
// [120, 255, 255, 255, 255,
255, 255]
}
~~~
~~~i3.d
// i3.d
void main ()
{
string s = "x";
static assert (is (typeof (s) == typeof ("x")));
assert (s == "x");
char [7] c7 = s; // throws RangeError
}
$ dmd -g i3
$ ./i3
core.exception.RangeError@i3.d(7): Range violation
??:? _d_arrayboundsp [...]
i3.d:7 _Dmain [...]
~~~
Shall I file them to bugzilla?
