Filter by value when reading and calculating data using pgfplotstable

Question

I am trying to plot data from a file, using pgfplots for the plot and pgfplotstable to read in the data and calculate a new column (cummulative values from another row), which is plotted as well. My data sometimes has bad values (negative values when all values should be non-negative) and I would like to filter those values from the plot and the calculation, either by ignoring them or setting them to zero. pgfplots has a convenient filter option for the plots, but this has no effect on the calculated column.

Here's a MWE of what I am trying to accomplish. There should not be any negative values in the plots. How should I filter the values?

\documentclass{minimal}

 \usepackage{filecontents}
 \begin{filecontents}{data.csv}


    Time,TemperatureC,DewpointC,PressurehPa,WindDirection,WindDirectionDegrees,WindSpeedKMH,WindSpeedGustKMH,Humidity,HourlyPrecipMM,Conditions,Clouds,dailyrainMM,SoftwareType,DateUTC,
2012-10-26 00:51:00,26.9,26.1,1009.4,SSE,163,8.0,22.5,95,0.0,,,0.0,weatherlink.com 1.10,2012-10-26 04:51:00, 
2012-10-26 01:06:00,26.9,26.1,1009.0,SW,231,8.0,20.9,94,0.0,,,0.0,weatherlink.com 1.10,2012-10-26 05:06:00, 
2012-10-26 01:21:00,26.9,26.1,1009.0,SSE,151,14.5,-1607.4,94,-2539.7,,,0.0,weatherlink.com 1.10,2012-10-26 05:21:00, 
2012-10-26 01:36:00,27.0,26.1,1009.0,SSE,155,4.8,22.5,93,0.0,,,0.0,weatherlink.com 1.10,2012-10-26 05:36:00, 
2012-10-26 01:51:00,27.1,26.1,1009.4,South,174,6.4,19.3,93,0.0,,,0.0,weatherlink.com 1.10,2012,
 \end{filecontents}

% UNITS
 \usepackage{siunitx}

% PGFPLOTS and TABLES
 \usepackage{tikz}
 \usepackage{pgfplots}
   \pgfplotsset{
     width=10cm,
     compat=1.7}
   \usepgfplotslibrary{dateplot}
   \usepgfplotslibrary{groupplots}

\usepackage{pgfplotstable}
   \pgfplotstableset{
     col sep=comma,
     create on use/SumP/.style={
       create col/expr={\pgfmathaccuma + \thisrow{HourlyPrecipMM}}
       }
   }
% ------------------------------------------------- Document starts here
\begin{document}

\pgfplotstableread{data.csv}\data



  \begin{tikzpicture}
    \begin{groupplot}[
      group style={%
        group size=1 by 2,
        horizontal sep=0pt,
        vertical sep=10pt,
        xticklabels at=edge bottom,
        xlabels at=edge bottom,
        ylabels at=edge left,
      },
      legend style={draw=none}, 
      legend pos=north west, 
      legend cell align=left, 
      date coordinates in=x, 
      xtick=,
      xticklabel style={rotate=90,anchor=near xticklabel},
      xticklabel=\month-\day. \hour:\minute, 
      date ZERO=2012-10-23, 
      minor tick num=12, 
      footnotesize,
      width=10cm,
      xlabel=Time / \si{\hour}, 
      ylabel=P / \si{\mm},
      ybar
      ]
      \nextgroupplot[]
      \addplot+[const plot mark left,fill=none] table[x=Time,y=HourlyPrecipMM]
      {\data};
      \addlegendentry{Rainfall depth}
      \addplot+[const plot mark left, fill=none,dashed] table[x=Time,y=SumP]
      {\data}; 
      \addlegendentry{Mass Rainfall}
      \nextgroupplot[ymax=100,height=4cm]
      \addplot+[const plot mark left,fill=none] table[x=Time,y=HourlyPrecipMM]
      {\data};
      \addlegendentry{Rainfall depth}

    \end{groupplot}
  \end{tikzpicture}

\end{document}

%%% Local Variables: 
%%% mode: latex
%%% TeX-engine: xetex
%%% TeX-master: t
%%% End: 

Answer 1

For accumulating only the non-negative values, you can use the expression \pgfmathaccuma + max(\thisrow{HourlyPrecipMM},0), where max(\thisrow{HourlyPrecipMM},0) evaluates to the data value if it's non-negative, and 0 if it's negative.

For filtering negative coordinates from the plots, you can set the key restrict y to domain=0: (leaving the upper bound empty).