Hi,

I have noticed that the spatial index performance of large MITAB generated files is much slower than that of an index generated by MapInfo and I wondered whether this is a known issue and whether there is a fix for the problem ?

A bit of background. I have a MapInfo dataset which contains about 110,000 polygons representing building outlines of New York and I need a quick way of querying which building is under the mouse cursor.

For a repeatible test, I've used OGRINFO.exe to perform the spatial query, but the same performance is observed when using the C++ MITAB library directly. The average time for 10 runs of this query with the MITab generated file is 3.6 seconds. However, the MapInfo generated file, performance is much better at 0.2s.

The command line I used was:

ogrinfo.exe -spat 588171 4515483 588171 4515483 ny.tab ny

I'm not sure whether this is a contributing factor, but the file sizes of the MITab and MapInfo generated vectors are different. The original source data was MIF/MID and converted it to TAB using TAB2TAB. This generated the following files:

 Directory of J:\MapInfoData

19/07/2006  20:06         6,737,350 NY.dat
19/07/2006  20:06           427,760 NY.id
19/07/2006  20:06        20,071,936 NY.ind
19/07/2006  20:06        13,559,808 NY.map
19/07/2006  20:06               190 NY.tab


If I then load those files into MapInfo and save a copy of the table as NY2, the corresponding file sizes become:

 Directory of J:\MapInfoData

19/07/2006  20:20         6,737,350 NY2.DAT
19/07/2006  20:20           427,760 NY2.ID
19/07/2006  20:20        15,512,064 NY2.IND
19/07/2006  20:20        16,565,760 NY2.MAP
19/07/2006  20:20               202 NY2.TAB


Using TAB2TAB again on these NY2 files, the filesizes revert back to those of the NY1 list - implying that MapInfo and MITab consider the contents to be the same, although they express the contents of MAP and IND files differently.

It would appear MITab is quite happy to quickly use the spatial indexes generated by MapInfo, but is unable to generate optimal spatial indexes itself.

These tests were conducted with MITAB 1.3 and the latest 1.5.0 release. I've also obtained the latest GDAL distribution which has a much newer OGL library and built a version of TAB2TAB but the performance was not improved.

I'd really appreciate comments from anyone who has observed similar spatial index performance issues.

Is the problem likely to be in MITab or OGR ?


Many thanks,

Andrew


__._,_.___


SPONSORED LINKS
Affiliate program software Construction software program Medical software program
Accounting software program Estate program real software Software development program


YAHOO! GROUPS LINKS




__,_._,___

Reply via email to