Sign in Join

Live chat for Free Pascal, Lazarus, Delphi and every Pascal programmer.

Mailing lists / fpc-pascal / 2018-12

precedence "is" and "and"

11 messages · 4 people · –

Reply by email Subscribe to this list Your subscription

Read-only archive. Use j and k to move between messages.

Mattias Gaertner

Hi,

According to the docs, the "is" operator is fourth level, the "and" is second level, so the "and" must be computed before the "is". But it seems fpc treats "is" and "and" as same level:

{$mode objfpc}
uses Classes;
var
  b: boolean;
  o: TObject;
begin
  // this compiles, but should fail:
  if o is TComponent and b then ;
  // this fails, correct
  if b and o is TComponent then ;
  // for completeness, this fails:
  if TComponent and b then ;
end.

The strange thing, is that the Delphi compiler works as fpc and the Delphi docs says the same as the fpc docs:
http://docwiki.embarcadero.com/RADStudio/Tokyo/en/Expressions_(Delphi)#Operator_Precedence

Mattias

Mattias Gaertner
to Mattias Gaertner
Show quoted text 5 lines
[...] According to the docs, the "is" operator is fourth level, the "and" is second level, so the "and" must be computed before the "is". But it seems fpc treats "is" and "and" as same level: [...]

Btw, there is a difference between fpc and delphi treating "is" and "or" precedence. The following compiles in fpc, but not in Delphi:

if b or o is TComponent then ;

Same with "xor".

This compiles in fpc and Delphi:
if o is TComponent or b then ;

It seems under Delphi "is", "and", "or", "xor" have the same precedence level and the Delphi docs are wrong.

Mattias

Mattias Gaertner
to Marco van de Voort
Show quoted text 6 lines
Op 2018-12-08 om 20:38 schreef Mattias Gaertner via fpc-pascal: > According to the docs, the "is" operator is fourth level, the "and" > is second level, so the "and" must be computed before the "is". But > it seems fpc treats "is" and "and" as same level: logical (boolean) vs bitwise AND difference?

I don't see where the bitwise should be involved here. There is no integer.
The docs do not mention a different precedence lvl for logical/bitwise "and".
It does not explain difference between "and" and "or".

Mattias

Michael Van Canneyt
to Mattias Gaertner
Mattias Gaertner wrote 7 lines
On Sat, 8 Dec 2018 21:23:57 +0100 Marco van de Voort wrote: > Op 2018-12-08 om 20:38 schreef Mattias Gaertner via fpc-pascal: > > According to the docs, the "is" operator is fourth level, the "and" > > is second level, so the "and" must be computed before the "is". But > > it seems fpc treats "is" and "and" as same level:

Hm.

In delphi, AS is second level, and 'is' is fourth level.

http://docwiki.embarcadero.com/RADStudio/Tokyo/en/Expressions_(Delphi)

If FPC does this differently, then I think this is an incompatibility.

Show quoted text 7 lines
> > logical (boolean) vs bitwise AND difference? I don't see where the bitwise should be involved here. There is no integer. The docs do not mention a different precedence lvl for logical/bitwise "and".

Because to the best of my knowledge, here isn't any difference in precedence, so there is nothing to explain.

It does not explain difference between "and" and "or".

Of course it does. See section on Logical operators

https://www.freepascal.org/docs-html/current/ref/refsu46.html

or boolean operators

https://www.freepascal.org/docs-html/current/ref/refsu47.html

I must confess it assumes that the reader knows what logical 'and' and 'or' means, but someone who does not know that will probably not read the docs. (although, he will also not understand "not" either and maybe read the docs...)

Michael.

Mattias Gaertner
to Michael Van Canneyt
Show quoted text 6 lines
[...] In delphi, AS is second level, and 'is' is fourth level. http://docwiki.embarcadero.com/RADStudio/Tokyo/en/Expressions_(Delphi) If FPC does this differently, then I think this is an incompatibility.

What has "as" to do with this thread?

Show quoted text 9 lines
>> logical (boolean) vs bitwise AND difference? > > I don't see where the bitwise should be involved here. There is no > integer. > The docs do not mention a different precedence lvl for > logical/bitwise "and". Because to the best of my knowledge, here isn't any difference in precedence, so there is nothing to explain.

Yes, that's my point. Even if bitwise "and" would play a role here, the precedence would still be the same.

Show quoted text 9 lines
> It does not explain difference between "and" and "or". Of course it does. See section on Logical operators https://www.freepascal.org/docs-html/current/ref/refsu46.html or boolean operators https://www.freepascal.org/docs-html/current/ref/refsu47.html

I looked at it, but I still don't understand.
Perhaps you can explain the examples I gave?

// compiles, but should fail:
if o is TComponent and b then ;
if o is TComponent or b then ;
// compiles in fpc, but not in Delphi:
if b or o is TComponent then ;
// does not compile, correct
if b and o is TComponent then ;

Mattias

Martin Wynne
to Mattias Gaertner: According to the docs, the "is" operator is fourth level, the "and" is second level, so…
Show quoted text 6 lines
var b: boolean; o: TObject; begin // this compiles, but should fail: if o is TComponent and b then ;

It will compile if $BOOLEVAL is on the default (-) because the result can be determined without considering precedence, see:

https://www.freepascal.org/docs-html/prog/progsu4.html

Try it the other way round and it will fail:

o: TComponent;
if o is TObject

Martin.

Michael Van Canneyt
to Mattias Gaertner: On Sat, 8 Dec 2018 22:03:11 +0100 (CET) What has "as" to do with this thread? Yes, that's…
Mattias Gaertner wrote 11 lines
On Sat, 8 Dec 2018 22:03:11 +0100 (CET) Michael Van Canneyt wrote: > [...] > In delphi, AS is second level, and 'is' is fourth level. > > http://docwiki.embarcadero.com/RADStudio/Tokyo/en/Expressions_(Delphi) > > If FPC does this differently, then I think this is an incompatibility. What has "as" to do with this thread?

Perhaps FPC treats "is" and "as" on the same level, it would explain the error ?

Show quoted text 14 lines
> >> logical (boolean) vs bitwise AND difference? > > > > I don't see where the bitwise should be involved here. There is no > > integer. > > The docs do not mention a different precedence lvl for > > logical/bitwise "and". > > Because to the best of my knowledge, here isn't any difference in > precedence, so there is nothing to explain. Yes, that's my point. Even if bitwise "and" would play a role here, the precedence would still be the same.

Agreed. It was Marco who brought this up ?

Show quoted text 18 lines
> > It does not explain difference between "and" and "or". > > Of course it does. See section on Logical operators > > https://www.freepascal.org/docs-html/current/ref/refsu46.html > > or boolean operators > > https://www.freepascal.org/docs-html/current/ref/refsu47.html I looked at it, but I still don't understand. Perhaps you can explain the examples I gave? // compiles, but should fail: if o is TComponent and b then ; if o is TComponent or b then ;

My hypothesis is that FPC treats 'is' and 'as' on level 2 (contrary to documentation). This is consistent with what you observe.

If 'is' is second level (i.e. the same as "as") then that would explain why it compiles, because the 'o is TComponent' will be evaluated first:

in the first line because evaluation happens from left to right, in the second line because 'is' takes precedence over 'or'.

// compiles in fpc, but not in Delphi: if b or o is TComponent then ;

Delphi is consistent with 'is' being on a lover level: "is" is lower level, so b or O is treated first.

Again, FPC's behaviour is consistent with treating 'is' on the same level as 'as': because 'is' takes precedence over 'or', you get no error.

// does not compile, correct if b and o is TComponent then ;

Indeed. (b and O) is evaluated first (left to right)

All is consistent with "is" and "as" being on the same level.

This is not how delphi does it.

So we adapt the documentation, or make the compiler delphi compatible.

Given that this has been in the compiler for years, I guess most people use brackets (I know I do)

Michael.

Marco van de Voort
to Michael Van Canneyt: Hm. In delphi, AS is second level, and 'is' is fourth level.…

Op 2018-12-08 om 22:03 schreef Michael Van Canneyt:

Show quoted text 6 lines
> I don't see where the bitwise should be involved here. There is no > integer. > The docs do not mention a different precedence lvl for logical/bitwise > "and".

My line of thought was the docs might mean bitwise there.

Because to the best of my knowledge, here isn't any difference in precedence, so there is nothing to explain.

And that was what my post was, checking that.

I'm more of the "more parenthesis never hurt anybody" school, so operator precedence beyond the normal mathematical is not my strong point.

Show quoted text 7 lines
I must confess it assumes that the reader knows what logical 'and' and 'or' means, but someone who does not know that will probably not read the docs. (although, he will also not understand "not" either and maybe read the docs...)

Maybe clarify that?

Michael Van Canneyt
to Marco van de Voort
Marco van de Voort wrote 8 lines
> Because to the best of my knowledge, here isn't any difference in > precedence, so there is nothing to explain. > And that was what my post was, checking that. I'm more of the "more parenthesis never hurt anybody" school, so operator precedence beyond the normal mathematical is not my strong point.

I am of the same school. When in doubt, use parentheses. This is unambiguous.

Show quoted text 8 lines
> I must confess it assumes that the reader knows what logical 'and' and > 'or' > means, but someone who does not know that will probably not read the > docs. > (although, he will also not understand "not" either and maybe read the > docs...) > Maybe clarify that?

Surely you are joking ?

The documentation is there to explain how the pascal language works, not how programming and computers in general work.

Michael.