# PGFPlots error “Dimension too large” when plotting .csv data

I did search on here for a solution and I did read the manual, but i couldn't find an answer (or at least one that I understood).

I am plotting some data in a .csv file and for some reason when the x axis value goes above 930.08 it gives the error "Dimension too large". What is strange, though, I have other data that does exceed 930.08 and that works fine, although it starts at 800, not 930 - this "erroneous" data is essentially a segment of larger data range but each sample is at a smaller samplerate (original data: start=800, increment=0.25, end= 1100, new data: start=930, increment: 0.01, end=960).

The data which gives the error is

a,b,c
930.00,0.0001095886,0.03875491
930.01,0.0001499818,0.0448156
930.02,0.0001117049,0.03977829
930.03,0.0001122457,0.03972992
930.04,0.0001595432,0.04589581
930.05,9.773631e-05,0.03843789
930.06,0.0001590169,0.04583743
930.07,0.000153542,0.04522526
930.08,9.953637e-05,0.03739852


The data before 930.08 doesn't give the errors.

The code I am using to plot the graph is:

\begin{center}
\begin{tikzpicture}
\begin{loglogaxis}[black,
width=14cm,
height=10cm,
mark=none,
xlabel=Scaling Factor,
transpose legend,
legend style={at={(0.5,-0.15)},anchor=north},
legend entries={MSE, Accuracy},
legend style={nodes=right}]

\addplot[no markers,blue] table [x=a, y=b, col sep=comma,] {
./chapter4/refined2b.csv
};
\addplot[no markers,red] table [x=a, y=c, col sep=comma,] {
./chapter4/refined2b.csv
};
\end{loglogaxis}
\end{tikzpicture}
\end{center}


Many thanks to anyone who responds.

• Welcome to TeX.sx! – Peter Jansson Apr 17 '13 at 19:56
• have you tried specifying ymax=<something>? – cmhughes Apr 17 '13 at 19:57
• Hi cmhughes, i tried doing "xmin=930,xmax=960,ymin=1e-7,ymax=2", which is roughly the range of coordinates, but this still gave the error. I think it has something to do with "axis equal", in the manual it says: "The axis equal key will be confused if x and y have a very different scale.", and in this plot this is certainly true. The default (false) gives the aforementioned error, if I change it to "true" the error disappears, but the plot is useless since all the plotted points are clumped together and you can't see them. – anderson Apr 17 '13 at 20:07
• Your x coordinate differences are too small to be distingished by a log plot. Using a semilogyaxis works without any problems. – percusse Apr 17 '13 at 20:08
• Thank you that fixed the problem (and explained it too), you've saved me a headache I really appreciate it. – anderson Apr 17 '13 at 20:12

Possibly having very little steps in the x coordinates forces pgfplots to the numerical accuracy limits. Instead using the logarithmic axes for y and linear for x works without any problem.

\begin{semilogyaxis}[black,
width=14cm,
height=10cm,
mark=none,
xlabel=Scaling Factor,
transpose legend,
legend style={at={(0.5,-0.15)},anchor=north},
legend entries={MSE, Accuracy},
legend style={nodes=right}]

\addplot[no markers,blue] table [x=a, y=b, col sep=comma,] {
./chapter4/refined2b.csv
};
\addplot[no markers,red] table [x=a, y=c, col sep=comma,] {
./chapter4/refined2b.csv
};
\end{semilogyaxis}