I don't think that this is a bug. I think that this is a consequence of the inexact nature of double precision numbers and the fact that you are trying create histograms where the data values and bin locations are the same. If you decrease the step size by a tiny amount, the correct result is returned. In addition, if you change a tiny amount to the minimum and maximum, the correct result will be returned. IDL> mx = 1.5d IDL> mn = 0.8d IDL> step = 0.23d IDL> arr = [mn:mx:step] IDL> hh = histogram(arr, binsize=step, loc=loc, min=mn-.0001d, max=mx+.0001d) IDL> print, hh 1 1 1 1 IDL> hh = histogram(arr, binsize=step-0.0000001d, loc=loc) IDL> print, hh 1 1 1 1 -David Starbuck Harris Geospatial Solutions

Yes, the "bug" is not exact wording. And your right, this is a consequence of the inexact nature of double precision numbers. The "bug" means problems of HISTOGRAM function algorithm. The HISTOGRAM function make a decide that the value of 1.03d( "arr[1]" in first writing) is smaller than 1.03d("loc[1]" in first writing). This is the problem. At least the example in the first writing, the result VALUE_LOCATE functin is correct(print, value_locate(loc, arr)). There is no problems. In the case of Python bridge. mn = 0.8d mx = 1.5d step = 0.23d arr = [mn:mx:step] h = histogram(arr, min=mn, max=mx, binsize=step, loc=loc) print, h ;; This is IDL histogram result 2 1 0 1 bins = [loc, loc[-1]] np = python.import('numpy') res = np.histogram(arr, bins=bins) print, res[0] ;; This Python result. 1, 1, 1, 1 The result of Python histogram function is correct. Thanks.