I use the following gnuplot script in order to plot a data file
reset
unset key
set size 1,1
set xrange [-10.1:11]
set yrange [-45:45]
set xlabel 'x'
set lmargin 6
set label 1 "~x{0.7.}" font "Helvetica, 20" at graph -0.1, graph 0.5
set xtics 2
set ytics 15
set mxtics 5
set mytics 5
plot "pss_data.dat" u 1:2 w dots lc rgb 'black'
set term postscript eps enhanced "Helvetica" 20 size 7in, 5in
set output 'plot.eps'
replot
reset
set terminal windows
quit
The exported .eps file is the following.
Well, in fact this is the corresponding .pdf file using Adobe Acrobat XI in order to make the conversion. However, the .pdf output contains not only the plot but all the unwanted black area above it! In an attempt to get rid off the white area I used the command line
epstopdf plot.eps
The output is the following
Now, the white area has been removed but the label at the y axis is also missing!
Any ideas? I want to have in a .pdf file only the plot (without the above white area) but with the label at the y axis.
Many thanks in advance.
Your bounding box may be incorrectly set. You can try using epstool on the eps you create:
epstool --bbox myeps.eps myneweps.eps
That should calculate the bounding box correctly, but give you a margin of zero. If you can't/don't want to install it, try adjusting the bounding box manually. There is a line near the top of the .eps file which looks like this:
%%BoundingBox: 50 50 554 770
The four numbers are the y offset, x offset, y max and x max of the output (in terms of margins you can think of them as top, left, bottom, right). You can try decreasing the second number (increasing the left margin) to see if that reveals your y axis label.
I would avoid the conversion all together by using one of gnuplot's pdf terminals (I like pdfcairo) and just use ylabel instead of set label 1 ... at graph.... Here's a simple script that you can modify for your purposes:
set term pdfcairo enhanced font "Helvetica,20"
set output "test.pdf"
set ylabel "~x{0.7.}" rotate by 0 #default rotation is 90
set xlabel "x"
plot sin(x)
Ultimately, what is happening with your script is that gnuplot is putting the label off of the viewable canvas. Some reason adobe still puts the label on the (converted) output, but I would assert that they are wrong in this case -- (they're essentially ignoring your bounding box). Of course, you could move/adjust the bounding box as suggested in the answer by andyras -- but I would argue that is a pretty hacky solution.
Related
Can you tell us how to make the chart with all headings and captions fit into the fixed vertical size and the horizontal size is floating?
If I set small fixed image sizes, some axes captions disappear, and if I set large fixed image sizes, then there is an empty space on the sides.
And I want you to be able to specify the height of the image and gnuplot will determine for itself what its width must be to fit everything you need.
For example (margin - empty space, but on the right, the inscription didn't fit:
set terminal pngcairo notransparent enhanced font "Calibri, 15" fontscale 1.0 size 800, 800; set zeroaxis;
Apparently, gnuplot does not automatically provide enough space for long numbers in the colorbox.
Since you have a logarithmic scale from 0.0000001 to 0.1, you might want to consider to display your numbers in a different format, i.e. from 10^-7 to 10^-1.
Code:
### no long numbers in colorbox
reset session
set colorbox
set palette rgb 33,13,10
set xrange [-7:-1]
set cbrange [1e-7:0.1]
set logscale cb
set format cb "10^{%T}"
plot '+' u 1:1:(10**$1) w l lw 3 palette z notitle
### end of code
Result:
I try to put the space to the index in the letters of the labels of Gnuplot. I use the following command:
set terminal pngcairo size 500,400 enhanced font 'Verdana,10'
set output 'test.png'
set xlabel 'atan(~{/Symbol w}{.6\~}_2)'
set ylabel 'atan(~{/Symbol w}{.6\~}_2)'
plot sin(x)
As you can see from the plot here the _2 over-plot the symbol. Could someone help me to understand how to put extra space to the indexes?
Admittedly, this is a bit of a hacky solution but it seems to work:
set xlabel 'atan(~{/Symbol w}{.6\~} _2)'
set ylabel 'atan(~{/Symbol w}{.6\~} _2)'
Adding the space after the symbol means that the subscript 2 doesn't overlap with it:
Okay I have been working with gnuplot for sometime and have one question. I cannot make the plot size equal to the terminal size I set it to. There are examples of setting
Lmargin,rmargin, tmargin and bmargin to 0 but this doesn't work on 'splot' which is 3d plotting. So I want to know what is the workaround for it?
For color maps you can use plot ... with image instead of set pm3d map if you want to use the margin options as usual. For 3d plots, as you say, the margin options are not available, and a workaround would be to shift the position of your graph and scale it. To do that, use set origin and set size respectively.
For instance, splot x*y yields the following:
If you want to reduce the margins while keeping the same overall terminal size, you can try:
set origin -0.1,-0.1
set size 1.2,1.2
splot x*y
which gives you:
You probably get the idea. Note I set a background color just so you can visualize where the margins lie because of the white background in the Stack Overflow website.
I am trying to create an adjacency matrix of a graph with gnuplot. The input graph has much more nodes than the pixels of the image that I will create. So, even for very sparse graphs I get very dark images.
My script is the following:
set terminal png size 800,800
set output "output.png"
set xrange []
set yrange [] reverse
set format y ""
set format x ""
unset xtics
unset x2tics
unset ytics
unset y2tics
plot 'data.dat' using 2:1 with dots lc rgb "#000000" notitle
I have thought of using 'every' to take a sample of the graph but I am not sure how I can choose an appropriate step. Using nodes/pixels for step creates a graph sparser than what I expected. Another idea is to use a heat map with colors making up for the shortage of pixels. However, I am not aware of how to create a heat map with gnuplot, since all the examples I find online seem to use a matrix for input while I use points for plotting.
Can someone provide me with an example for a heat map with points instead of a matrix as input or justify which step would be appropriate for use in order to get a good representation?
I'm having difficulty getting GNUplot to properly render some of my data. Basically I have some 2D heightmap data (for example 512 x 128 data points) and would like to be able to plot this in a "pixel perfect" fashion (i.e. one pixel per data point). Unfortunately at the moment, I'm getting artefacts (see inset) due to the number of pixels being slightly larger than the number of data points. I know I can set the total size of the output using:
set terminal png size 800,400
but is there a way to set just the plot area size in pixels?
Thanks!
From the help of the lua/TikZ terminal (there is an option plotsize) it is suggested to remove the plot margins so that plot fills the whole canvas.
set lmargin 0
set rmargin 0
set tmargin 0
set bmargin 0
set terminal png size 512,128
set out 'out.png'
# a checkerboard pattern to test the plot results for errors
f(x,y)=( (x<=0) || (x>=63) || (y<=0) || (y>=63) ) ? 0.5 : ((x+y)%2==0)?0:1
# the 'map' file contains only coordinates.
plot[-1:63][0:64] 'map' u 1:2:(f($1,$2)) with image
unset out
However, this has the drawbacks that you will not have tic labels and so on. You can of course calculate values for the margins yourself, so that they match with the plot size in the end: use
set lmargin at screen 0.25
for example.
Anyway, I would also recommend going for vector graphics whenever possible.