Plotting from two data sets delimited two different ways - gnuplot

Question

I need to plot data from a .csv file and from a white space separated file. Both sets of data need to appear on the same plot.
data1.dat
#t y
1 1
2 1
3 1
and
data2.csv
#t,y
1,2
2,2
3,2
normally I would do the following if both were .csv sets:
set datafile separator ','
plot 'data1.csv' using 1:2,'data2.csv' using 1:2
Is there some way to include the setting of the separation character in the plot statement?
plot 'data1.dat' using 1:2,'data2.csv' using datafile separator ',' using 1:2
The above does not work and I tried many different variations of the above code....I had no luck.

Answer 1

You can give more than one character to set datafile separator, in your case ", ". All these are then individually treated as separators. (A tab caracter can be given as "\t", you need to put double quotes around the string then!)
$dat << EOD
1 2,4
2 2,5
3 1,6
4 4,4
EOD
set xr [0.5:4.5]
set dataf sep ", "
plot $dat us 1:2:3 w yerrorbars
Note that the explicit separator characters each count as exactly one separator. "4, 4" with set dataf sep ", " evaluates to "three columns, second is a missing value. If you have incompatible formats in one plot, you can import the data for each subplot with its own separators set using set table $<name>. (check "help datablocks")
If your data files have a very difficult format:
gnuplots using specifier accepts a libC scanf() format string
plot "-" us 1:2 "%lf,%lf"
1,2
2,3
3,4
e
You can give a different format string for every file on your plot command. Note that gnuplot only accepts "double" fp numbers for input, so you have to use the %le or %lf specifier.
Check help using examples, and here is a full description of the format.

Answer 2

AFAIK, there isn't a way to specify the separator. However, if you're in a POSIX compliant environment (and your gnuplot supports pipes -- which most do), you can farm the work out to awk pretty easily:
plot 'data1.dat' using 1:2,\
"<awk -F, '{print $1,$2}' data2.csv" using 1:2

Answer 3

Not just for "retro"-fun, but also for current gnuplot versions, I guess this is the only(?) gnuplot-only solution which works with all versions even with versions before the time of OP's question.
The "trick" is: if you set datafile separator "," and read the first (and only) stringcolumn (of a whitespace separated file), i.e. strcol(1) will contain the full line. Now, you can simply split the string with word() and convert it into a floating point number with real().
If your original data has at least one space after the comma,
1, -0.2
2, -0.1
3, 0.0
the data would be plotted correctly with keeping the separator as whitespace, since the comma after the first column's data will be ignored during number interpretation.
Although, for newer gnuplot versions (>=4.6.7, Apr 2015) you have the possibility to define several separators, however, which will not work as you might think, because
set datafile separator ", "
will interpret each space as column separator. So, if you have an undefined and variable number of spaces your plot command would fail.
Anyway, here is the "always" working solution:
Data:
SO14262760_1.dat (with variable number of spaces)
1 -0.1
2 0.0
3 +0.1
SO14262760_2.dat (with no or some space after ,)
1,-0.2
2,-0.1
3, 0.0
Script: (works with gnuplot>=4.4.0, March 2010)
### different column separators in two files with one plot command
reset
FILE1 = "SO14262760_1.dat"
FILE2 = "SO14262760_2.dat"
set datafile separator ","
myCol(n) = real(word(strcol(1),n))
plot FILE1 u (myCol(1)):(myCol(2)) w lp pt 7 lc rgb "red", \
FILE2 u 1:2 w lp pt 7 lc rgb "blue"
### end of script
Result: