[ncl-talk] Issue with creating a bar chart plot

Rick Brownrigg brownrig at ucar.edu
Tue Jun 9 07:39:06 MDT 2020 

(back online)

Hi Andrew,

The errors have to do indirectly with the reference line; If I remove those
resources, I get a histogram without error messages being thrown. I was
able to track precisely where in the code things go awry, however I do not
understand the logic leading up to that point, so I really don't know what
is the real problem. I unfortunately do not have time to dig into this at
the level it would require to understand this code (oh for an interactive
ncl debugger).

There were some changes to gsn_csm.ncl that dealt with reference lines
around the 2017-2018 time frame, but they seem to be minor and not directly
related to the code where the miscalculation of box widths occurs, so
again, its hard to tell.

I wish I had a better answer
Rick

On Tue, Jun 9, 2020 at 5:46 AM Andrew Kren - NOAA Affiliate <
andrew.kren at noaa.gov> wrote:

> Hi Rick,
>
> Here is a functioning code. I just saved the plotting section and made my
> own test array, giving the same error. I know I don't need to convert to
> integer in my min and max calculation but that shouldn't matter.
>
> Thanks,
>
> On Mon, Jun 8, 2020 at 10:05 PM Rick Brownrigg <brownrig at ucar.edu> wrote:
>
>> Offline for the moment -- Andrew if you can send me the data and a
>> functionally complete script, I'll try to dig into it deeper.
>>
>> On Mon, Jun 8, 2020 at 6:31 PM Andrew Kren - NOAA Affiliate <
>> andrew.kren at noaa.gov> wrote:
>>
>>> Hi ncl-talk, Dave, and Rick,
>>>
>>> I took both of your guys' advice and here is what I found. I first
>>> removed tolong but that did not fix it. I then tried dong rounding, as well
>>> as not doing any conversion, but that still resulted in the error. Then I
>>> tried commenting out the defining of the bar chart width, but it still came
>>> up with the same error:
>>>
>>> (0) get_bar_arrays: Fatal: The array is not monotonically increasing.
>>>
>>> (0) Cannot draw a bar chart.
>>>
>>> fatal:Illegal right-hand side type for assignment
>>>
>>>
>>> Here is what I have for my plot section:
>>>
>>>
>>> outfile = "fcst_efuuyr3wd6_top_1_" + (i)
>>>
>>>
>>> length1 = temperature_top(i,0:23)
>>>
>>>
>>>  wks = gsn_open_wks("eps",outfile)  ; send graphics to PNG file
>>>
>>>  res = True
>>>
>>>  res at gsnScale = True
>>>
>>>  res at gsnFrame = False
>>>
>>>  res at gsnDraw = True
>>>
>>> ; these four resources allow the user to stretch the plot size, and
>>>
>>> ; decide exactly where on the page to draw it.
>>>
>>>
>>>  ;res at gsnMaximize = True
>>>
>>>   res at vpXF = 0.01            ; In page coordinates, where to start
>>>
>>>   res at vpYF = 0.9  ; 0.75            ; the plot
>>>
>>>   res at vpHeightF = 0.27            ; Changes the aspect ratio
>>>
>>>   res at vpWidthF  = 0.97
>>>
>>>
>>>  x_axis = ispan(4,73,3)
>>>
>>>  x_axis at _FillValue = 999
>>>
>>>
>>>
>>>   res at tiMainString  = names_14ers(i) + " 3-Day Forecast (Peak: " +
>>> hgts_14ers_str(i) + " ft)"
>>>
>>>   res at gsnXYBarChart = True            ; Create bar plot
>>>
>>>   res at tiMainOffsetYF = 0.006
>>>
>>>   res at tiMainFontHeightF = 0.02
>>>
>>>  res at tmXTMode       = "Explicit" ; Define own tick mark labels.
>>>
>>>  res at tmXTValues     = x_axis
>>>
>>>  res at tmXTOn         = False             ; turn on the top tickmarks
>>>
>>>  res at tmXBOn         = False
>>>
>>>  res at tmYROn         = False
>>>
>>>  res at tmYRMinorOn    = False
>>>
>>>  res at tmYLOn         = False
>>>
>>>  res at tmYLMinorOn    = False
>>>
>>>  res at tmXTMinorOn    = False
>>>
>>>  res at gsnXRefLine = (/26.5,50.5/)
>>>
>>>  res at gsnXRefLineColor = (/"gray40","gray40"/)
>>>
>>>  res at gsnXRefLineDashPatterns = (/2,2/)
>>>
>>>  res at gsnXRefLineThicknesses = (/1.05,1.05/)
>>>
>>>  res at tiYAxisOffsetXF = 0.035
>>>
>>>  res at tiYAxisString = "Temperature"
>>>
>>>
>>>   res at tiYAxisFontHeightF = 0.014
>>>
>>>
>>> range_temp = ( tofloat( max(tolong(temperature_top(i,0:23))) -
>>> min(tolong(temperature_top(i,0:23)))))
>>>
>>> max_temp = tofloat(max(tolong(temperature_top(i,0:23)))) +
>>> (tofloat(max(tolong(temperature_top(i,0:23))) -
>>> min(tolong(temperature_top(i,0:23))))) * 0.8
>>>
>>> min_temp = tofloat(min(tolong(temperature_top(i,0:23)))) -
>>> (tofloat(max(tolong(temperature_top(i,0:23))) -
>>> min(tolong(temperature_top(i,0:23))))) * 0.1
>>>
>>>
>>>  res at trYMaxF         = max_temp   ; max value on y-axis
>>>
>>>  res at trYMinF         = min_temp
>>>
>>> ; res at gsnXYBarChartBarWidth = 0.75                 ; change bar widths
>>> 0.75 originally
>>>
>>>  res at gsnXYBarChartColors = "Orange"
>>>
>>>
>>>  ; draw 5 separate temperature plots so can make figure big
>>>
>>>  res at trXMinF               = 0                    ; adds space on
>>> either end
>>>
>>>  res at trXMaxF               =  75                    ; of the 1st and
>>> last bars
>>>
>>>
>>> plot = gsn_csm_xy(wks,x_axis,temperature_top(i,0:23),res)
>>>
>>>
>>> My x_axis values are: 4 to 73 by 3
>>>
>>> My temperature_top values are:
>>>
>>>
>>> (0) 14
>>>
>>> (1) 11
>>>
>>> (2) 11
>>>
>>> (3) 12
>>>
>>> (4) 14
>>>
>>> (5) 21
>>>
>>> (6) 24
>>>
>>> (7) 23
>>>
>>> (8) 23
>>>
>>> (9) 22
>>>
>>> (10) 23
>>>
>>> (11) 26
>>>
>>> (12) 33
>>>
>>> (13) 36
>>>
>>> (14) 33
>>>
>>> (15) 28
>>>
>>> (16) 25
>>>
>>> (17) 25
>>>
>>> (18) 25
>>>
>>> (19) 29
>>>
>>> (20) 38
>>>
>>> (21) 41
>>>
>>> (22) 39
>>>
>>> (23) 35
>>>
>>>
>>> I don't understand why the error could still be happening.
>>>
>>>
>>> Thanks,
>>>
>>> On Mon, Jun 8, 2020 at 1:40 PM Dave Allured - NOAA Affiliate <
>>> dave.allured at noaa.gov> wrote:
>>>
>>>> Andrew,
>>>>
>>>> In the face of unknown behavior, I would try a couple things.  First,
>>>> go back to your original code without my previous suggestion of reversing
>>>> the axes.  That was based on a wrong assumption.
>>>>
>>>> 1.  I do not understand the purpose of "tolong" here.  Data type "long"
>>>> is normally used only for special purposes in NCL, and there is a chance it
>>>> is causing trouble.  Also, "tolong" would reduce plot accuracy by integer
>>>> truncation.  Try simply removing "tolong", and use temperature_top(i,0:23)
>>>> directly in the function call.  Tofloat, todouble, or toint would be safer
>>>> alternatives if a conversion function is really needed.  If you really want
>>>> integer rounding here, then you should use the round function to reduce
>>>> roundoff error.
>>>>
>>>> 2.  Try completely isolating the plot function from the rest of the
>>>> program.  Write the X and Y arrays to a small Netcdf file.  Print (res) and
>>>> manually recode the printed results into new NCL code lines in a small new
>>>> test program.  Omit all "res" attributes that are not relevant for this
>>>> plot type.  Read the Netcdf file, then call open_wks and gsn_csm_xy, and
>>>> see if the problem remains.  This way, at least you end up with a small
>>>> reproducer and small data that you can send to the mailing list.  HTH.
>>>>
>>>>
>>>> On Mon, Jun 8, 2020 at 7:17 AM Rick Brownrigg <brownrig at ucar.edu>
>>>> wrote:
>>>>
>>>>> I took a quick look at the code. As the errors indicate, it's
>>>>> happening in a function call get_bart_width(). The calculation of the
>>>>> variable "delta" is clearly wrong, and that sets off the warning, and
>>>>> triggers the error later in get_bar_arrays about non-monotonicity (it's
>>>>> actually not that the array is non-monotonic, it's that it has negative
>>>>> widths).
>>>>>
>>>>> Beyond that, I can not tell why the calculation is wrong, particularly
>>>>> without the data. That part of the code hasn't changed in some years, so
>>>>> the v6.5.0 vs v6.6.2 difference is a mystery.
>>>>>
>>>>> FWIW...
>>>>> Rick
>>>>>
>>>>>
>>>>> On Sun, Jun 7, 2020 at 6:58 PM Andrew Kren - NOAA Affiliate via
>>>>> ncl-talk <ncl-talk at mailman.ucar.edu> wrote:
>>>>>
>>>>>> I looked into these things. My x_axis is simply created this way:
>>>>>>
>>>>>> x_axis = ispan(4,73,3)
>>>>>>
>>>>>>
>>>>>> And the printed values are correct, increasing by 3 starting at 4, 7,
>>>>>> etc.
>>>>>>
>>>>>>
>>>>>> So this is still kinda confusing as to why.
>>>>>>
>>>>>> On Sun, Jun 7, 2020 at 5:41 PM Dave Allured - NOAA Affiliate <
>>>>>> dave.allured at noaa.gov> wrote:
>>>>>>
>>>>>>> It is only the horizontal axis that needs to be monotonically
>>>>>>> increasing.  That is probably what the error message is about.  Naturally,
>>>>>>> the vertical access may fluctuate anywhere.  You might want to check the
>>>>>>> values of x_axis in your code to see if they make sense.
>>>>>>>
>>>>>>> Meanwhile, you should be able to insert a quick fix for x_axis with
>>>>>>> temporary arrays, something like this.  This assumes that x_axis is
>>>>>>> strictly decreasing rather than increasing.  If you change the order, then
>>>>>>> you must also change the order for the vertical axis data, in the same
>>>>>>> way.  X and Y values are supposed to form ordered pairs to make the bar
>>>>>>> chart.
>>>>>>>
>>>>>>> xplot = x_axis(::-1)
>>>>>>> yplot = tolong (temperature_top(i,23:0:-1))
>>>>>>> plot = gsn_csm_xy (wks, xplot, yplot, res)
>>>>>>>
>>>>>>> I have no idea what changed to cause a newer NCL version to change
>>>>>>> from the original behavior.
>>>>>>>
>>>>>>>
>>>>>>> On Sun, Jun 7, 2020 at 1:57 PM Andrew Kren - NOAA Affiliate via
>>>>>>> ncl-talk <ncl-talk at mailman.ucar.edu> wrote:
>>>>>>>
>>>>>>>> Dear ncl-talk,
>>>>>>>>
>>>>>>>> I upgraded my NCL to the latest version. Previously I was using
>>>>>>>> version 6.5. When I upgraded, one of my ncl codes no longer runs
>>>>>>>> successfully but creates an error code. I am creating a bar chart with
>>>>>>>> gsn_csm_xy. I originally did not get these errors so I am unsure what could
>>>>>>>> be going on. I tried to debug the issue but to no avail. Here are my errors:
>>>>>>>>
>>>>>>>> (0) get_bar_widths: Warning: The bar width(s) you selected (0.75)
>>>>>>>> is larger than the smallest delta (-11).
>>>>>>>>
>>>>>>>> (0) Defaulting to -11.
>>>>>>>>
>>>>>>>> (0) get_bar_arrays: Fatal: The array is not monotonically
>>>>>>>> increasing.
>>>>>>>>
>>>>>>>> (0) Cannot draw a bar chart.
>>>>>>>>
>>>>>>>> fatal:Illegal right-hand side type for assignment
>>>>>>>>
>>>>>>>>
>>>>>>>> I checked that my array I am plotting has all correct values, and
>>>>>>>> it does. I didn't know a bar chart had to have monotonically increasing
>>>>>>>> values. Is there a way around this?
>>>>>>>>
>>>>>>>>
>>>>>>>> The code where it fails is below, in the call to plot.
>>>>>>>>
>>>>>>>>
>>>>>>>> plot = gsn_csm_xy(wks,x_axis,tolong(temperature_top(i,0:23)),res)
>>>>>>>>
>>>>>>>> Thanks,
>>>>>>>>
>>>>>>>> --
>>>>>>>> Andrew Kren
>>>>>>>> Meteorologist
>>>>>>>> US Dept of Commerce
>>>>>>>> National Oceanic and Atmospheric Administration
>>>>>>>> National Weather Service Raleigh, NC
>>>>>>>> 1005 Capability Drive, Suite 300
>>>>>>>> Centennial Campus
>>>>>>>> Raleigh, NC 27606-5226
>>>>>>>>
>>>>>>>
>>>
>>> --
>>> Andrew Kren
>>> Meteorologist
>>> US Dept of Commerce
>>> National Oceanic and Atmospheric Administration
>>> National Weather Service Raleigh, NC
>>> 1005 Capability Drive, Suite 300
>>> Centennial Campus
>>> Raleigh, NC 27606-5226
>>>
>>>
>
> --
> Andrew Kren
> Meteorologist
> US Dept of Commerce
> National Oceanic and Atmospheric Administration
> National Weather Service Raleigh, NC
> 1005 Capability Drive, Suite 300
> Centennial Campus
> Raleigh, NC 27606-5226
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.ucar.edu/pipermail/ncl-talk/attachments/20200609/ed5e608d/attachment.html>

More information about the ncl-talk mailing list