www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Idea: Common errors messages

reply Dawid =?ISO-8859-2?Q?Ci=EA=BFarkiewicz?= <arael fov.pl> writes:
I like D very much, but when I'm writing code I spend too much time thinking
about "what does this error message means". This problem is common for a
lot of compilers, but D - as a rather new language - makes me even more
confused sometimes because I'm not used to it. 

Let's say I've just:

foreach(int b, _bogus) {
    b=1;
}

and DMD gives me:

: no identifier for declarator

I've started to look for typos in naming and such when actually I've used
',' where ';' was needed. I know this is my fault and this mistake means
lack of wisdom and itss rather easy to find. But it would be perfect if D
compiler could give hints for such common mistakes. It wouldn't interrupt
or even slower normal parsing and similar. DMD could just take line with
error at the end of parsing process and compare it with some regular
expresions. It could be even a command line switch -ne (noob errors) or
something like this.

What do you think?
-- 
Dawid Ciężarkiewicz | arael
jid: arael fov.pl
Dec 16 2004
next sibling parent reply =?ISO-8859-2?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
Dawid Ciężarkiewicz wrote:


 Let's say I've just:
 
 foreach(int b, _bogus) {
     b=1;
 }
 
 and DMD gives me:
 
 : no identifier for declarator


Funny, the GDC compiler instead says: (v0.8)


  found ')' when expecting ';'


(both should be pointing to the line, which
some IDEs can use to highlight the error...)


Better error messages could come later, at this
point it is mostly 1) fixing D bugs 2) giving errors
in the first place, instead of throwing assertions :-)

But some more sample code, with suggestions for better
error messages than the current ones, would be nice ?

--anders
Dec 16 2004
parent reply Dawid =?ISO-8859-2?Q?Ci=EA=BFarkiewicz?= <arael fov.pl> writes:
Anders F Björklund wrote:


 (both should be pointing to the line, which
 some IDEs can use to highlight the error...)


DMD is pointing to the line - I've just erased the line number that was
right before ':'.


 Better error messages could come later, at this
 point it is mostly 1) fixing D bugs 2) giving errors
 in the first place, instead of throwing assertions :-)


Yes. There is many more important things. :)


 But some more sample code, with suggestions for better
 error messages than the current ones, would be nice ?


I'll try to give some examples soon. The problem with finding examples of
"stupid mistakes" is that you have to make such a mistake first. But it's
rather easy for me so... :D



For now the "you forgot the ';' after previous line" and checking the
keywords environment (for commas, {} blocks, ";" and such) is what comes to
my mind. Maybe there is too less such a places when this whole idea could
be used. But maybe someone else was thinking about something similar
before, when he was looking at his code after many hours of work and
couldn't find "so simple mistake" for a long time.
-- 
Dawid Ciężarkiewicz | arael
jid: arael fov.pl
Dec 16 2004
parent reply "Asaf Karagila" <kas1 netvision.net.il> writes:
the problem i had with "null is not an lvalue" is exactly
that kind of message. i had no clue what went wrong.
only by accident i came to see that it's the unneeded
"const" declaration that caused the error.

i agree, better error messages would produce healthier
code, and as well would help newbies to learn the language..

- Asaf.

"Dawid Ciężarkiewicz" <arael fov.pl> wrote in message 
news:cprmf1$1p3a$1 digitaldaemon.com...

 Anders F Björklund wrote:


 (both should be pointing to the line, which
 some IDEs can use to highlight the error...)


 DMD is pointing to the line - I've just erased the line number that was
 right before ':'.


 Better error messages could come later, at this
 point it is mostly 1) fixing D bugs 2) giving errors
 in the first place, instead of throwing assertions :-)


 Yes. There is many more important things. :)


 But some more sample code, with suggestions for better
 error messages than the current ones, would be nice ?


 I'll try to give some examples soon. The problem with finding examples of
 "stupid mistakes" is that you have to make such a mistake first. But it's
 rather easy for me so... :D



 For now the "you forgot the ';' after previous line" and checking the
 keywords environment (for commas, {} blocks, ";" and such) is what comes 
 to
 my mind. Maybe there is too less such a places when this whole idea could
 be used. But maybe someone else was thinking about something similar
 before, when he was looking at his code after many hours of work and
 couldn't find "so simple mistake" for a long time.
 -- 
 Dawid Ciężarkiewicz | arael
 jid: arael fov.pl
Dec 17 2004
parent reply Georg Wrede <Georg_member pathlink.com> writes:
In article <cpunfm$25s5$1 digitaldaemon.com>, Asaf Karagila says...

the problem i had with "null is not an lvalue" is exactly
that kind of message. i had no clue what went wrong.
only by accident i came to see that it's the unneeded
"const" declaration that caused the error.


Yes, although exceptional in D, this particular example reminds
me of C++ compiler error messages: even in the unlikely 
event that they happen to refer to the right line number,
chances are that the subject of the lament is off by miles.


i agree, better error messages would produce healthier
code, and as well would help newbies to learn the language..


Not only that. Established D programmers would waste less
time debugging, newbies would learn the language a lot faster,
Real Gurus would find their (rare trivial) mistakes faster.


 Better error messages could come later, at this
 point it is mostly 1) fixing D bugs 2) giving errors
 in the first place, instead of throwing assertions :-)






Agreed. But at 1.0 we really have to have a compiler that gives
unambiguous error messages, hits the line, and the messages
will be (A LOT) more informative than "null is not an lvalue".

Actually something like a "-newbie" compiler switch might not
be such a bad idea. This should not imply cheesy patronizing,
rather it should be aimed towards an average (professional?)
programmer, who just happens to be new to D.

Two things multiply the rate of expansion of D: (1) how fast 
people get confident with the compiler, and (2) how fast a 
programmer "outputs" valid, bug and error free code. 
(Equals Productivity!)

Score 0.5 in both, and you've quartered the success of D.


 But some more sample code, with suggestions for better
 error messages than the current ones, would be nice ?






This might be one perfect thing to load off Walter's back!?


 I'll try to give some examples soon. The problem with finding examples of
 "stupid mistakes" is that you have to make such a mistake first. But it's
 rather easy for me so... :D




Actually, a job that even a D-newbie (who is not pathetically over-
worked, like most of us here) could do, and thence make a major 
contribution to D -- would be to take (say) the wc-example in the docs,
and just deliberately make typos (like omitting a ";", misspelling a
keyword, skipping or using an extra parenthesis somewhere), change
types, declarations, etc. etc.

The ensuing error messages could be explained in such an article.

Ideally this should be written by somebody who's not used D for too
long, because then one tends to become "home blind" and forget what
the really hard issues, and tough parts were.
Dec 18 2004
parent J C Calvarese <jcc7 cox.net> writes:
Georg Wrede wrote:

 In article <cpunfm$25s5$1 digitaldaemon.com>, Asaf Karagila says...
 


...

Better error messages could come later, at this
point it is mostly 1) fixing D bugs 2) giving errors
in the first place, instead of throwing assertions :-)






 
 
 Agreed. But at 1.0 we really have to have a compiler that gives
 unambiguous error messages, hits the line, and the messages
 will be (A LOT) more informative than "null is not an lvalue".


Since I want D 1.0 release sometime /this decade/, it'd be acceptable to 
me if "null is not an lvalue" isn't improved until later. Perhaps you 
meant to say "at 10.0"?

What I've tried to do when I've found what I thought was a particularly 
horrible error message is to post the message (and code to produce it 
and maybe a suggestion) to the bugs newsgroup. I don't know if Walter 
plans on making my suggested improvements, but I at least feel better 
for registering the complaint.

-- 
Justin (a/k/a jcc7)
http://jcc_7.tripod.com/d/
Dec 18 2004
prev sibling next sibling parent Thomas Kuehne <thomas-dloop kuehne.thisisspam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dawid schrieb am Thu, 16 Dec 2004 10:04:51 +0100:

 I like D very much, but when I'm writing code I spend too much time thinking
 about "what does this error message means". This problem is common for a
 lot of compilers, but D - as a rather new language - makes me even more
 confused sometimes because I'm not used to it.

 Let's say I've just:

 foreach(int b, _bogus) {
     b=1;
 }

 and DMD gives me:

: no identifier for declarator

 I've started to look for typos in naming and such when actually I've used
 ',' where ';' was needed.


Enhancing "lexer.c" and the various "::error(const char *format, ...)"
should allow the compiler to print the line number and the position within
the line that caused the error.

By adding a rescan function to the various "::error(const char *format, ...)"
it might be possible to print the actual content of the line and mark
the trouble spot.

Potential problem: 
- - Lexer::peek(Token *ct)
- - trans-module errors


 I know this is my fault and this mistake means
 lack of wisdom and itss rather easy to find. But it would be perfect if D
 compiler could give hints for such common mistakes. It wouldn't interrupt
 or even slower normal parsing and similar. DMD could just take line with
 error at the end of parsing process and compare it with some regular
 expresions. It could be even a command line switch -ne (noob errors) or
 something like this.


I'm unsure about your "regular expression" suggestion.

Surely, the error messages could be improved, e.g.
"no identifier for declarator '_bogus'" instead of
"no identifier for declarator".

Thomas


-----BEGIN PGP SIGNATURE-----

iD8DBQFBwVxV3w+/yD4P9tIRApRNAJ9f+6JA0qoik6oEWOpwYib8V1gh1ACgil5K
om4vbBkLxEXLr5EswyH9gds=
=qC08
-----END PGP SIGNATURE-----
Dec 16 2004
prev sibling next sibling parent J C Calvarese <jcc7 cox.net> writes:
In article <cprj3m$1l8g$1 digitaldaemon.com>, Dawid
=?ISO-8859-2?Q?Ci=EA=BFarkiewicz?= says...

I like D very much, but when I'm writing code I spend too much time thinking
about "what does this error message means". This problem is common for a
lot of compilers, but D - as a rather new language - makes me even more
confused sometimes because I'm not used to it. 


While we're waiting for Walter to fix the bugs, we can make on own list. Here's
a start: http://www.prowiki.org/wiki4d/wiki.cgi?ErrorMessages

jcc7
Dec 16 2004
prev sibling parent reply "Walter" <newshound digitalmars.com> writes:
"Dawid Ciężarkiewicz" <arael fov.pl> wrote in message
news:cprj3m$1l8g$1 digitaldaemon.com...

 I like D very much, but when I'm writing code I spend too much time


thinking

 about "what does this error message means". This problem is common for a
 lot of compilers, but D - as a rather new language - makes me even more
 confused sometimes because I'm not used to it.

 Let's say I've just:

 foreach(int b, _bogus) {
     b=1;
 }

 and DMD gives me:

 : no identifier for declarator

 I've started to look for typos in naming and such when actually I've used
 ',' where ';' was needed. I know this is my fault and this mistake means
 lack of wisdom and itss rather easy to find. But it would be perfect if D
 compiler could give hints for such common mistakes. It wouldn't interrupt
 or even slower normal parsing and similar. DMD could just take line with
 error at the end of parsing process and compare it with some regular
 expresions. It could be even a command line switch -ne (noob errors) or
 something like this.

 What do you think?


I think you're right. D error messages do need improvement, and they are
gradually getting better.
Dec 16 2004
parent reply "Simon Buchan" <currently no.where> writes:
On Thu, 16 Dec 2004 15:12:17 -0800, Walter <newshound digitalmars.com>  
wrote:


 "Dawid Ciê¿arkiewicz" <arael fov.pl> wrote in message
 news:cprj3m$1l8g$1 digitaldaemon.com...

 I like D very much, but when I'm writing code I spend too much time


 thinking

 about "what does this error message means". This problem is common for a
 lot of compilers, but D - as a rather new language - makes me even more
 confused sometimes because I'm not used to it.

 Let's say I've just:

 foreach(int b, _bogus) {
     b=1;
 }

 and DMD gives me:

 : no identifier for declarator

 I've started to look for typos in naming and such when actually I've  
 used
 ',' where ';' was needed. I know this is my fault and this mistake means
 lack of wisdom and itss rather easy to find. But it would be perfect if  
 D
 compiler could give hints for such common mistakes. It wouldn't  
 interrupt
 or even slower normal parsing and similar. DMD could just take line with
 error at the end of parsing process and compare it with some regular
 expresions. It could be even a command line switch -ne (noob errors) or
 something like this.

 What do you think?


 I think you're right. D error messages do need improvement, and they are
 gradually getting better.


Walter, Am I correct in my assumption that it is currently not possible for
an end-user to build DMD? It might help in cases like this a bit if people  
can
actually make it work better, and see that they didn't screw up. (I know  
(DMC/
legal issues?) you don't want to distro the backend source, but maybe it
precompiled as a lib or something?)

-- 
"Unhappy Microsoft customers have a funny way of becoming Linux,
Salesforce.com and Oracle customers." - www.microsoft-watch.com:
"The Year in Review: Microsoft Opens Up"
--
"I plan on at least one critical patch every month, and I haven't been  
disappointed."
- Adam Hansen, manager of security at Sonnenschein Nath & Rosenthal LLP
(Quote from http://www.eweek.com/article2/0,1759,1736104,00.asp)
--
"It's been a challenge to "reteach or retrain" Web users to pay for  
content, said Pizey"
-Wired website: "The Incredible Shrinking Comic"
Dec 21 2004
parent Anonymous Coward <Anonymous_member pathlink.com> writes:
In article <opsjduhir8jccy7t simon.mshome.net>, Simon Buchan says...

Walter, Am I correct in my assumption that it is currently not possible for
an end-user to build DMD? It might help in cases like this a bit if people  
can
actually make it work better, and see that they didn't screw up. (I know  
(DMC/
legal issues?) you don't want to distro the backend source, but maybe it
precompiled as a lib or something?)


How about GDC?
Dec 22 2004