Here's how to setup an attribute map in GTL using the example you listed earlier:
ods path WORK.TEMPLAT(UPDATE) SASHELP.TMPLMST (READ);
proc template;
define statgraph symbolchange;
begingraph;
discreteattrmap name="attrs" / ignorecase=true DISCRETELEGENDENTRYPOLICY=ATTRMAP;
value 'F' / lineattrs=(color=red pattern=1) markerattrs=(symbol=triangle color=blue);
value 'M' / lineattrs=(color=blue pattern=2) markerattrs=(symbol=triangle color=red);
enddiscreteattrmap;
discreteattrvar attrvar=_attrs_ var=sex attrmap="attrs" ;
layout overlay;
seriesplot y=height x=weight / group=_attrs_ name='a';
scatterplot y=height x=weight / group=_attrs_ name='b';
mergedlegend 'a' 'b'/ location=inside autoalign=(bottom);
endlayout;
endgraph;
end;
run;
ods graphics / reset;
proc sgrender data=sashelp.class template=symbolchange;
run;
The discreteattrmap block is similar to a proc format section, but for each value you list out you can set the individual style elements (markerattrs, lineattrs, etc.). You then setup a temporary variable with discreteattrvar that gets mapped to the variable that contains the values, and use that temporary variable in a GROUP option in your graph statements.
If you need the style to be based on two different variables (for lines and markers) there are two options. First is to make two different attribute maps (named differently) and assign them to your series and scatter plot statements individually. The second option is to make a new variable in your dataset with the unique combinations of the two variables (e.g. A and B = 1, A and C=2, B and C=3) and use the new variable in the attribute map.
In the above example I used a merged legend to combine the series and scatter plot legends together to get the appropriate legend that shows lines and markers together. There's supposed to be a way that you can directly feed an attribute map into a legend but I haven't been able to get it to work so there could be a bug there.
