Hello everyone,
We are currently using Solr 8.7 facets search to render a server generated
heatmap.
Our issue is the following :
Solr often return this kind of exception =>
"java.lang.IllegalArgumentException: Too many cells (743 x 261) for level 5
shape
Rect(minX=-18.610839843750004,maxX=13.996582031250002,minY=40.53050177574321,maxY=51.97134580885172)"
Here the facet json query was =>
{
"coord": {
"type": "heatmap",
"field": "coord_fieldname",
"geom": "[\"-18.610839843750004 40.53050177574321\" TO \"13.996582031250002
51.97134580885172\"]",
"distErrPct":0.025
}
}
The same query works if we put the distErrPct at 0.03.
We found that these lines of code, in
org.apache.lucene.spatial.prefix.HeatmapFacetCounter class were generating this
exception:
45 public static final int MAX_ROWS_OR_COLUMNS = (int)
Math.sqrt(ArrayUtil.MAX_ARRAY_LENGTH);
......
140 if (columns > MAX_ROWS_OR_COLUMNS || rows > MAX_ROWS_OR_COLUMNS || columns
* rows > maxCells) {
141 throw new IllegalArgumentException(
142 "Too many cells ("
143 + columns
144 + " x "
145 + rows
146 + ") for level "
147 + facetLevel
148 + " shape "
149 + inputRect);
150 }
Althought we don"t really understand how these maximum rows and columns counts
are calculated.
We would like to obtain maximum precision when calculating our heatmap.
Should we adapt distErrPct before each query in order to avoid this kind of
response from Solr ? (if yes, how so ?)
Or is there something else we are missing ?
Because Solr succeed in processing this query for most of our "geom" rectangles
in parameter, but it fails for some of them. The rectangles we send always have
same proportions, just the "zoom" level differs.
Thanks for your answers !
