Sorting according to more than one key in pgfplotstable

Question

Is it possible to sort rows of a pgfplotstable according to multiple keys?

For example, input file is as follows:

Mtx Kind P   Sp
A   LP   16  4.2
C   OPT  16  72.5
A   LP   64  20.3
B   OPT  16  5.7
B   OPT  64  16.4
A   LP   256 90.4

We want to sort by column Kind first, then by Mtx, then by P. Desired table should look like as follows:

Mtx Kind P   Sp
B   OPT  16  5.7
B   OPT  64  16.4
C   OPT  16  72.5
A   LP   16  4.2
A   LP   64  20.3
A   LP   256 90.4    

I am able to sort more than one numeric column. However, since, I was not able to convert a string to a numeric value (e.g. via using hash of string), I could not sort such data. It is also ok to sort the file directly using multiple keys without using pgfplotstable.

Answer 1

Pgfplots comes with builtin support for single sort keys.

However, if is possible to plug in multiple sort keys -- with a bit of efford and inside knowledge:

\documentclass[a4paper]{standalone}

\usepackage{pgfplots}

\usepackage{pgfplotstable}

\begin{document}

\def\pgfplotsinvokeiflessthan#1#2#3#4{%
    \pgfkeysvalueof{/pgfplots/iflessthan/.@cmd}{#1}{#2}{#3}{#4}\pgfeov
}%
\def\pgfplotsmulticmpthree#1#2#3#4#5#6\do#7#8{%
    \pgfplotsset{string <}%
    \pgfplotsinvokeiflessthan{#1}{#4}{%
        % first key <:
        #7%
    }{%
        \pgfplotsinvokeiflessthan{#4}{#1}{%
            % first key >:
            #8%
        }{%
            % first key ==:
            \pgfplotsset{string <}%
            \pgfplotsinvokeiflessthan{#2}{#5}{%
                % second key <
                #7%
            }{%
                \pgfplotsinvokeiflessthan{#5}{#2}{%
                    % second key >
                    #8%
                }{%
                    % second key ==
                    \pgfplotsset{float <}%
                    \pgfplotsinvokeiflessthan{#3}{#6}{%
                        % third key <
                        #7%
                    }{%
                        % third key >=
                        #8%
                    }%
                }%
            }%
        }%
    }%
}%

\pgfplotstabletypeset[
    create on use/sortkey/.style={
        create col/assign/.code={%
            \edef\entry{{\thisrow{Kind}}{\thisrow{Mtx}}{\thisrow{P}}}%
            \pgfkeyslet{/pgfplots/table/create col/next content}\entry
        }
    },
    sort key=sortkey,
    sort cmp={%
        iflessthan/.code args={#1#2#3#4}{%
            \edef\temp{#1#2}%
            \expandafter\pgfplotsmulticmpthree\temp\do{#3}{#4}%
        },
    },
    sort,
    columns/Mtx/.style={string type},
    columns/Kind/.style={string type},
]{
Mtx Kind P   Sp
A   LP   16  4.2
C   OPT  16  72.5
A   LP   64  20.3
B   OPT  16  5.7
B   OPT  64  16.4
A   LP   256 90.4
}
\end{document}

The approach consists of two parts: (1) the creation of a temporary column called sortkey which expands to {A}{LP}{16} for the first row and (2) the definition of a customized sort cmp. The customized sort cmp here uses a multi-key comparison (lexicographical compare) with string < for the first two keys and float < for the third.

Note that sort key=sortkey instantiates the create on use specification while sort cmp defines the customized comparison.

The key sort activates the sort functionality

The approach is fairly general and should probably become a general-purpose utility; I will think about a suitable predefined style.