Martin,

Are you doing any kind of geometry validation on the parser? It would be good 
to log geometry errors and all that too. Invalid polygons present a real 
problem and are hard to detect in some cases.

Adam

> On Jun 16, 2015, at 5:12 AM, Martin Desruisseaux 
> <[email protected]> wrote:
> 
> Hello all
> 
> In the process of reading a Well Known Text (WKT), some non-fatal
> warnings may happen (this email ignores fatal errors, which throw
> exception without logging anything). Examples of warnings are:
> 
>  * A metadata depends on an optional module which is not on the
>    classpath (example: TimeExtent).
>  * An unknown keyword has been found. The WKT specification requires
>    that we ignore unknown keywords (in order to allow users to extend
>    the WKT with their own elements, for example triaxal ellipsoids
>    [1]), but suggests that we emit a warning.
>  * An inconsistency is found. The specification said "In the event of
>    conflict in values given in the CRS WKT string and given by an
>    authority through an object’s name or an identifier, reading
>    software should throw an exception or give users a warning message.
>    The WKT values should be assumed to prevail." [2]
> 
> An example of warning message currently produced by the WKT parser is
> (advises about rewording the warning are welcome):
> 
>    Parsing of “WGS 84” done, but some elements were ignored.
>     • The text contains unknown elements:
>        ‣ “MYEXT” in PRIMEM, GEOGCS.
> 
> What should be the policy about non-fatal warnings in Apache SIS? The
> current approach is:
> 
>  * When using the WKTFormat object (which give maximum control on WKT
>    parsing and formatting), no warnings are logged. But a getWarnings()
>    method is provided, which allow users to decide what they want to do
>    [3].
>  * When using the CRS.fromWKT(String) static convenience method [4],
>    currently warnings are ignored.
> 
> Should non-fatal warnings be logged when using a static convenience
> method? If yes, should it be at Level.WARNING or at another lever?
> 
>    Martin
> 
> 
> [1] http://docs.opengeospatial.org/is/12-063r5/12-063r5.html#146
> [2] http://docs.opengeospatial.org/is/12-063r5/12-063r5.html#139
> [3] 
> https://builds.apache.org/job/sis-dev/javadoc/org/apache/sis/io/wkt/Warnings.html
> [4] 
> https://builds.apache.org/job/sis-dev/javadoc/org/apache/sis/referencing/CRS.html#fromWKT-java.lang.String-
> 
> 

Reply via email to