0005159: wrong location after line directive

(0005679)
doligez (manager)
2010-10-12 11:46
edited on: 2010-10-12 11:46

What makes you think it should be 0 ? The regular compiler does the same thing as camlp4.
These offsets are from the start of the lexbuf, not from the start of an embedded (and possibly nonexistent) file.

(0005681)
Hendrik Tews (reporter)
2010-10-12 23:08

The documentation says that Loc.start_off "Returns the number of
characters from the begining of the file of the begining of this
location" and because file_name="b.ml" and in b.ml the let is at
position 0 start_off should return 0.

(0005682)
doligez (manager)
2010-10-13 16:03

Hm. The documentation needs to be rewritten in a less ambiguous way.
In this case, "the file" is your two-line file that contains the # and the let
(i.e. the real file that OCaml is parsing).

In any case, I don't think you should rely on anything else than the
difference between offset and bol.

(0005683)
Hendrik Tews (reporter)
2010-10-18 09:56

doligez wrote:

    Hm. The documentation needs to be rewritten in a less ambiguous way.

I don't see any ambiguity, there is only one file, which name is
returned by Loc.file_name.

    In any case, I don't think you should rely on anything else than the
    difference between offset and bol.

You mean file_name and start/stop_line may return invalid values?
Why are these functions present then?

How would you report an error location with only the difference
between offset and bol? Something like

  File unknown, line unknown, characters 8-9:
  Error: ...

?

(0005685)
doligez (manager)
2010-10-19 14:09

Hendrik Tews wrote:

  I don't see any ambiguity, there is only one file, which name is
  returned by Loc.file_name.

Actually, there is only one file, the one that OCaml is parsing (your two-line
file that starts with '# 1 "b.ml"'). The fact that you're interpreting it differently
definitely indicates ambiguity.

  You mean file_name and start/stop_line may return invalid values?
  Why are these functions present then?

No, I mean as far as offset and bol are concerned, only the difference is
significant. Of course you can use the file name and line number.

(0006910)
doligez (manager)
2012-02-10 10:08

Fixed by rewording the documentation to refer to the lexbuf rather than "the file".

Fixed in trunk [3.13.0] (commit 12144).

(0006911)
Hendrik Tews (reporter)
2012-02-10 10:40

May I suggest to update the documentation in camlp4/Camlp4/Sig.ml
as well? There Loc.start_off is still specified to return the
number of characters from the beginning of the file.

I would further suggest to add a note to camlp4/Camlp4/Sig.ml stating:

Note that character numbers in locations refer to character
numbers in the parsed character stream, while line numbers refer
to line numbers in the source file. The source file and the
parsed character stream differ, for instance, when the parsed
character stream contains a line number directive. The line
number directive will only update the file-name field and the
line-number field of the position. It makes therefore no sense to
use character numbers with the source file if the sources contain
line number directives.

(0006918)
doligez (manager)
2012-02-13 18:46

Done in trunk [3.13.0] (commit 12150)

Issue History
Date Modified	Username	Field	Change
2010-09-29 15:06	Hendrik Tews	New Issue
2010-09-29 15:06	Hendrik Tews	Status	new => assigned
2010-09-29 15:06	Hendrik Tews	Assigned To	=> ertai
2010-10-12 11:46	doligez	Note Added: 0005679
2010-10-12 11:46	doligez	Note Edited: 0005679
2010-10-12 23:08	Hendrik Tews	Note Added: 0005681
2010-10-13 16:03	doligez	Note Added: 0005682
2010-10-13 16:03	doligez	Assigned To	ertai => doligez
2010-10-18 09:56	Hendrik Tews	Note Added: 0005683
2010-10-19 14:09	doligez	Note Added: 0005685
2012-02-10 10:08	doligez	Note Added: 0006910
2012-02-10 10:08	doligez	Status	assigned => closed
2012-02-10 10:08	doligez	Resolution	open => fixed
2012-02-10 10:08	doligez	Fixed in Version	=> 3.13.0+dev
2012-02-10 10:40	Hendrik Tews	Note Added: 0006911
2012-02-10 10:40	Hendrik Tews	Status	closed => feedback
2012-02-10 10:40	Hendrik Tews	Resolution	fixed => reopened
2012-02-13 18:46	doligez	Note Added: 0006918
2012-02-13 18:46	doligez	Status	feedback => resolved
2012-02-13 18:46	doligez	Resolution	reopened => fixed

Notes
(0005679) doligez (manager) 2010-10-12 11:46 edited on: 2010-10-12 11:46	What makes you think it should be 0 ? The regular compiler does the same thing as camlp4. These offsets are from the start of the lexbuf, not from the start of an embedded (and possibly nonexistent) file.

(0005681) Hendrik Tews (reporter) 2010-10-12 23:08	The documentation says that Loc.start_off "Returns the number of characters from the begining of the file of the begining of this location" and because file_name="b.ml" and in b.ml the let is at position 0 start_off should return 0.

(0005682) doligez (manager) 2010-10-13 16:03	Hm. The documentation needs to be rewritten in a less ambiguous way. In this case, "the file" is your two-line file that contains the # and the let (i.e. the real file that OCaml is parsing). In any case, I don't think you should rely on anything else than the difference between offset and bol.

(0005683) Hendrik Tews (reporter) 2010-10-18 09:56	doligez wrote: Hm. The documentation needs to be rewritten in a less ambiguous way. I don't see any ambiguity, there is only one file, which name is returned by Loc.file_name. In any case, I don't think you should rely on anything else than the difference between offset and bol. You mean file_name and start/stop_line may return invalid values? Why are these functions present then? How would you report an error location with only the difference between offset and bol? Something like File unknown, line unknown, characters 8-9: Error: ... ?

(0005685) doligez (manager) 2010-10-19 14:09	Hendrik Tews wrote: I don't see any ambiguity, there is only one file, which name is returned by Loc.file_name. Actually, there is only one file, the one that OCaml is parsing (your two-line file that starts with '# 1 "b.ml"'). The fact that you're interpreting it differently definitely indicates ambiguity. You mean file_name and start/stop_line may return invalid values? Why are these functions present then? No, I mean as far as offset and bol are concerned, only the difference is significant. Of course you can use the file name and line number.

(0006910) doligez (manager) 2012-02-10 10:08	Fixed by rewording the documentation to refer to the lexbuf rather than "the file". Fixed in trunk [3.13.0] (commit 12144).

(0006911) Hendrik Tews (reporter) 2012-02-10 10:40	May I suggest to update the documentation in camlp4/Camlp4/Sig.ml as well? There Loc.start_off is still specified to return the number of characters from the beginning of the file. I would further suggest to add a note to camlp4/Camlp4/Sig.ml stating: Note that character numbers in locations refer to character numbers in the parsed character stream, while line numbers refer to line numbers in the source file. The source file and the parsed character stream differ, for instance, when the parsed character stream contains a line number directive. The line number directive will only update the file-name field and the line-number field of the position. It makes therefore no sense to use character numbers with the source file if the sources contain line number directives.

(0006918) doligez (manager) 2012-02-13 18:46	Done in trunk [3.13.0] (commit 12150)

Relationships