Handling of DBF NULL values

[unicoletti]

The dbf code in mapserver does not handle correctly null numeric values (and possibly other types too, though I haven't checked) and returns which is the internal dbf representation for a null value. This has undesired side effects on filters and leads to parse errors which are translated in exceptions in java mapscript eventually blocking the generation of the map.

As an example implementation see the DBFIsAttributeNULL() in shapelib/dbfopen.c

[unicoletti]

I suggest the following change and volunteer to help:

modify msDBFGetValueList so that it returns the "NULL" string for null values, then modify the lexer to handle the special NULL string. NULL would be a new keyword.

[sdlime]

What do other data sources do with NULLs? I don't think shapefiles are alone with this. Would be nice to be consistent across all providers.

Steve

[unicoletti]

Replying to sdlime:

What do other data sources do with NULLs? I don't think shapefiles are alone with this. Would be nice to be consistent across all providers.

Steve

Postgis seems to return an empty string (as that's the default behaviour of PQgetvalue) in all cases (numbers, strings, etc). Have not cheked other providers. As a quick fix we could convert null strings in empty strings and null numbers in 0. This would leave the parser untouched and expression evaluation will always work. This one and #2123 would be a nice addition to 5.0.

[sdlime]

Might check with the SDE and Oracle Spatial leads too. I agree, would be a nice addition... Would you do the impementation?

Steve

[unicoletti]

Replying to sdlime:

Might check with the SDE and Oracle Spatial leads too. I agree, would be a nice addition... Would you do the impementation?

For DBF, yes. I will send a cumulative patch for this and #2123.

[unicoletti]

Cumulative patch for 2123 and 2141

[unicoletti]

The patch works for me. I'll delay the commit until the segfault caused by msAddLabel is fixed.

[unicoletti]

Fixed in revision 6277.