QueryByAttributes

[dstrevinas]

I came across a problem while using python mapscript's queryByAttribute function. I have a layer containing the fieldname "STATE" amongst others. Issuing:

try:
   layer.queryByAttribute(map,"STA","AZ",MS_MULTIPLE)
except:
   print "attribute error"
   pass

By now the mapserver exception is caught for the faulty attribute. If a draw function is now issued:

map.draw()

The following error occurs:

<class '_mapscript.MapServerError'>: msDrawMap(): Image handling error. Failed to draw layer named 'countries'. msDBFGetItemIndex(): DBASE file error. Item 'STA' not found.

I reproduced it many times and in many occasions. I tried to clear any result cache but in vain. Shouldn;t mapserver handle the situation more logically?

[dmorissette]

I'm not setup to test with Python MapScript, but it seems to me that calling msResetErrorList() in the exception handling block might do the trick.

Closing for now. Please try adding a call to msResetErrorList() and if that doesn't work then please reopen and we'll reassign to one of the Python guys.

[dstrevinas]

At last i managed to get through the source and realized that QueryByAttributes sets a FILTERITEM within the layer object. If a map is drawn and the filteritem is not cleared, then mapserver draw functions complain (when the layer is queried with false fieldname). So the manual solution is:

try:
   layer.queryByAttribute(map,"STA","AZ",MS_MULTIPLE)
except:
   '''layer.filteritem=""'''
   print "attribute error"
   pass

Something that will confuse mapscript users. I didn't modify anything in source. Wait for any comment.

[dstrevinas]

Finally, setting filteritem to doesn't either work. The problem is that msQueryByAttributes in mapquery.c does not restore the previous state of the filters whenever it returns with an error condition. I create a function called msRestoreOldFilter and call it whenever msQueryByAttributes returns resolving the problem successfully. I attached a modified mapquery.c

[dmorissette]

The patch seems to make sense, but I still can't reproduce this with my own script and mapfile, so I am hesitant to apply a patch that I can't test.

Can you please share your layer definition? Perhaps that will give me some hints on what I may be missing in my testcase to reproduce this.

[dmorissette]

I am using PHP MapScript whose queryByAttributes() implementation is identical to the Python one, and it always cleans up the filter properly, so I wonder if it could be something related to the Python exception handling. Steve, Howard, does the try/except let the underlying C function finish normally or does it abort it?

[dmorissette]

Jim sent me a copy of his mapfile and data and I was able to reproduce the error with PHP (and determine that I had made a stupid error in my initial test script).

I have tested and committed the patch with a minor change to remove unnecessary strdup/frees on the old_filteritem and old_filter buffers in msRestoreOldFilter() (those existed before the patch).

Committed in trunk (future v6.0) r8837 and branch-5-4 (future 5.4.0) r8838.