[erlang-bugs] Parser bug: if missing ", " in list between records, the first is ignored, and no parser warnings are given.
Roland Karlsson
roland.karlsson@REDACTED
Sun Nov 9 12:19:55 CET 2014
I think you are mixing two things here.
You want a warning for the usage of #rec{}#rec{}.
Mattias wanted a warning when "data was lost".
Those are two different reasons for a warning.
I answered Mattias in my post.
Generally Erlang do not warn you if you have underscore
variables or anonymous terms (in expressions) that are
later not used. So - the warning Mattias wanted do break
common rules in the Erlang compilator.
The general #rec{}#rec{} warning is a bit more complicated.
Just writing it is maybe a bit strange and maybe a warning
might be in place. But - the macro usage I gave is quite legit I
think and not so horrible at all. And ... I do not know if a warning
that depends on it being a macro or not is a clever idea.
/Roland
On Sun, 09 Nov 2014 12:49:41 +0200
Loïc Hoguin <essen@REDACTED> wrote:
>
>
> On 11/09/2014 12:40 PM, Roland Karlsson wrote:
> >
> > On Sat, 8 Nov 2014 09:02:08 +0100
> > Mattias Waldau <mattias.waldau@REDACTED> wrote:
> >
> >> Yes, I can agree that
> >>
> >> #rec{a=2}#rec{b=3}
> >>
> >> is a matter of opinion. However, I completely fail to understand how
> >>
> >> #rec{a=2}#rec{a=3}
> >>
> >> can be a matter of opinion, since data in the code is LOST.
> >
> > Data is not lost.
> > The code creates a record rec with the field a being 2.
> > Then the code creates a copy of that record where the field now is 3.
> > The first record is instance not used though. But, as it has no name
> > that usually do not give a warning in Erlang, just like unused variables
> > with underscore names do not give a warning.
> >
> > But ... I can see why it might maybe be a warning there.
> > It IS a matter of opinion though IMHO.
>
> When I said it needs a warning earlier in the thread, I meant one you can set optionally, similar to +warn_export_all. Regardless of the contents of the record. You have to program in a very specific way to be annoyed by that kind of warning though.
>
> > You have to be aware of the following code though.
> >
> > (?MY_DEF_REC)#rec{a=3}
> >
> > Lets say the macro MY_DEF_REC creates a record rec with some
> > special default values. Lets say that one of the fields you have a
> > special default value for is a. Do you want a warning?
>
> And that would be one such way... though my heart would skip a beat seeing code like this.
>
> Most people simply do not write code that ends up as #rec{...}#rec{...}, and could very much benefit from a warning that avoids this error. Others can simply choose not to enable it.
>
> > /Roland
> >
> >
> >
> >
> >
> >
> >
> >
> >> There should be a warning. My sample below shows how catastropic the missing "," can be.
> >>
> >> /mattias
> >>
> >> On 07/11/2014 13:22, Roland Karlsson wrote:
> >>> Maybe we shall not discuss this for ever :)
> >>> It is a matter of opinion after all.
> >>>
> >>> But - it is the same IMHO. It is a matter of operators and the need
> >>> to set paranthesis when the precedents and right/left association is
> >>> not obvious.
> >>>
> >>> The rec example can be written thus
> >>>
> >>> (#rec{a=2})#rec{b=3}
> >>>
> >>> Then everything is clear.
> >>>
> >>>
> >>> /Roland
> >>>
> >>>
> >>>
> >>>> I'm not sure how that relates. This expression can be reasonably written
> >>>> on purpose. The same cannot be said for #rec{a=2}#rec{b=3}.
> >>>>
> >>>> On 11/07/2014 02:01 PM, Roland Karlsson wrote:
> >>>>> If there is a warning there, then all expressions including operators,
> >>>>> e.g. arithmetic, should have warnings. How about 10 + 1 rem 3 + 7. ?
> >>>>> Do you know the precedence of + and rem ? What is the result?
> >>>>>
> >>>>> /Roland
> >>>>>
> >>>>>
> >>>>>
> >>>>>> Warnings are ordinary Erlang code which is prone to errors.
> >>>>>>
> >>>>>> On 11/07/2014 01:46 PM, Roland Karlsson wrote:
> >>>>>>> No warning is needed methinks. It is just ordinary Erlang code.
> >>>>>>> It is equal to.
> >>>>>>>
> >>>>>>>> REC1 = #rec{b=3}.
> >>>>>>>> REC2 = REC1#rec{a=2}.
> >>>>>>> Nothing strange at all.
> >>>>>>>
> >>>>>>>
> >>>>>>> /Roland
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>> It is legal but there probably should be a warning for this
> >>>>>>>> particular
> >>>>>>>> case. I cannot think of a valid reason to do #rec{a=2}#rec{b=3} when
> >>>>>>>> you
> >>>>>>>> can just do #rec{a=2, b=3}.
> >>>>>>>>
> >>>>>>>> On 11/07/2014 10:09 AM, Vlad Dumitrescu wrote:
> >>>>>>>>> Hi Mattias,
> >>>>>>>>>
> >>>>>>>>> This isn't a bug. The expression #rec{a=2}#rec{b=3} is legal, and is
> >>>>>>>>> the
> >>>>>>>>> same thing as (#rec{a=2})#rec{b=3} or #rec{a=2,b=3}. If the keys
> >>>>>>>>> overlap, then the last one "wins".
> >>>>>>>>>
> >>>>>>>>> best regards,
> >>>>>>>>> Vlad
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> On Thu, Nov 6, 2014 at 9:21 AM, Mattias Waldau
> >>>>>>>>> <mattias.waldau@REDACTED <mailto:mattias.waldau@REDACTED>>
> >>>>>>>>> wrote:
> >>>>>>>>>
> >>>>>>>>> This is the program
> >>>>>>>>>
> >>>>>>>>> -module('missing-comma-not-detected').
> >>>>>>>>>
> >>>>>>>>> -export([data/1]).
> >>>>>>>>>
> >>>>>>>>> -compile(export_all).
> >>>>>>>>>
> >>>>>>>>> -record(shcnode, {
> >>>>>>>>> nodename :: any(),
> >>>>>>>>> y_axis :: number(),
> >>>>>>>>> children :: [any()],
> >>>>>>>>> article_number :: number(),
> >>>>>>>>> attribute_condition :: string()}).
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> data(1) ->
> >>>>>>>>> [
> >>>>>>>>> #shcnode{nodename = start, y_axis = 0, children =
> >>>>>>>>> [d1285]},
> >>>>>>>>> #shcnode{nodename = d1285, y_axis = 295, children =
> >>>>>>>>> [1939,
> >>>>>>>>> 1940]}
> >>>>>>>>> #shcnode{nodename = 1940, y_axis = 525, children =
> >>>>>>>>> [bm3203],
> >>>>>>>>> article_number = 14432861},
> >>>>>>>>> #shcnode{nodename = 1939, y_axis = 530, children =
> >>>>>>>>> [bm3203],
> >>>>>>>>> article_number = 14432860},
> >>>>>>>>> #shcnode{nodename = bm3203, y_axis = 915, children = []}
> >>>>>>>>> ].
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Note the missing "," after [1939, 1940]}
> >>>>>>>>>
> >>>>>>>>> Only 4 rows are detected
> >>>>>>>>>
> >>>>>>>>> length('missing-comma-not-detected':data(1)).
> >>>>>>>>> 4
> >>>>>>>>>
> >>>>>>>>> ---
> >>>>>>>>>
> >>>>>>>>> If I add the comma, I get
> >>>>>>>>>
> >>>>>>>>> length('missing-comma-not-detected':data(1)).
> >>>>>>>>> 5
> >>>>>>>>>
> >>>>>>>>> ----
> >>>>>>>>>
> >>>>>>>>> I am using Erlang 17.3 on ubuntu.
> >>>>>>>>>
> >>>>>>>>> erlang:system_info(otp_release).
> >>>>>>>>> "17"
> >>>>>>>>>
> >>>>>>>>> 36> |erl -V
> >>>>>>>>> Eshell V6.2 (abort with ^G)
> >>>>>>>>> 1> |
> >>>>>>>>>
> >>>>>>>>> ---
> >>>>>>>>>
> >>>>>>>>> Thanks,
> >>>>>>>>>
> >>>>>>>>> Mattias
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> _______________________________________________
> >>>>>>>>> erlang-bugs mailing list
> >>>>>>>>> erlang-bugs@REDACTED <mailto:erlang-bugs@REDACTED>
> >>>>>>>>> http://erlang.org/mailman/listinfo/erlang-bugs
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> _______________________________________________
> >>>>>>>>> erlang-bugs mailing list
> >>>>>>>>> erlang-bugs@REDACTED
> >>>>>>>>> http://erlang.org/mailman/listinfo/erlang-bugs
> >>>>>>>>>
> >>>>>>>> --
> >>>>>>>> Loïc Hoguin
> >>>>>>>> http://ninenines.eu
> >>>>>>>> _______________________________________________
> >>>>>>>> erlang-bugs mailing list
> >>>>>>>> erlang-bugs@REDACTED
> >>>>>>>> http://erlang.org/mailman/listinfo/erlang-bugs
> >>>>>>>>
> >>>>>>>
> >>>>>> --
> >>>>>> Loïc Hoguin
> >>>>>> http://ninenines.eu
> >>>>>>
> >>>>>
> >>>> --
> >>>> Loïc Hoguin
> >>>> http://ninenines.eu
> >>>> _______________________________________________
> >>>> erlang-bugs mailing list
> >>>> erlang-bugs@REDACTED
> >>>> http://erlang.org/mailman/listinfo/erlang-bugs
> >>>>
> >>>
> >>
> >> _______________________________________________
> >> erlang-bugs mailing list
> >> erlang-bugs@REDACTED
> >> http://erlang.org/mailman/listinfo/erlang-bugs
> >
>
> -- Loïc Hoguin
> http://ninenines.eu
--
Roland Karlsson <roland.karlsson@REDACTED>
More information about the erlang-bugs
mailing list