How to Plot CSV Columns with pgfplotsinvokeforeach

To plot several columns from one comma-separated file, read the file once as a PGFPlots table, then loop over the names of the columns you want to use as y-values. In the loop, #1 stands for the current column name. This is a convenient use of \pgfplotsinvokeforeach because its body can contain PGFPlots commands such as \addplot.

Complete example

The example below creates a small CSV file, reads it, and plots both runA and runB against the shared time column. Save it as a .tex file and compile it with a current LaTeX installation.

\documentclass{article}

\begin{filecontents*}{data.csv}
time,runA,runB
0,1,2
1,2.2,1.8
2,3.1,3.4
3,2.8,4.1
\end{filecontents*}

\usepackage{pgfplots}
\usepackage{pgfplotstable}
\pgfplotsset{compat=1.18}

\begin{document}

\pgfplotstableread[col sep=comma]{data.csv}\datatable

\begin{tikzpicture}
\begin{axis}[
  xlabel={Time},
  ylabel={Value},
  legend pos=north west,
  grid=major,
]
  \pgfplotsinvokeforeach{runA,runB}{%
    \addplot+[mark=*] table[x=time,y=#1] {\datatable};
    \addlegendentry{#1}
  }
\end{axis}
\end{tikzpicture}

\end{document}

The filecontents* environment writes data.csv when the document is compiled. If you already have a CSV file, remove that environment and keep the \pgfplotstableread command, changing data.csv to the file’s name or path.

How the loop works

\pgfplotstableread[col sep=comma]{data.csv}\datatable loads the table and stores it in the macro \datatable. The option col sep=comma tells PGFPlots that commas separate fields. The first row supplies the column names: in this example, time, runA, and runB.

The loop receives a comma-separated list of column names:

\pgfplotsinvokeforeach{runA,runB}{...}

On the first pass, #1 is runA; on the second, it is runB. The plot command therefore behaves as if you had written these two commands individually:

\addplot+[mark=*] table[x=time,y=runA] {\datatable};
\addplot+[mark=*] table[x=time,y=runB] {\datatable};

The x=time option selects the horizontal coordinate, while y=#1 selects the current vertical-coordinate column. Both series use the same x-column. The \addlegendentry{#1} line is inside the loop, so each plot receives its own legend entry. Remove that line if you do not need a legend, or replace #1 with a descriptive label if the CSV column names are not suitable for display.

Adapting the column list

To plot different y-columns, change only the list in \pgfplotsinvokeforeach. For example, {control,treatment,reference} will generate three plots, provided those names match columns in the table. To change the x-column, edit x=time to the appropriate header, such as x=distance.

For a file with many columns, you can keep the x-column fixed and list only the columns you want to plot. You do not need to include the x-column in the loop. If you need a different label for each line, write explicit labels in the loop body or define a separate mapping between column names and legend text; #1 itself is just the current list item.

Common mistakes

• Forgetting the delimiter option: PGFPlots commonly expects whitespace-separated data by default. For a comma-separated file, set col sep=comma when reading the table.
• Using a header that does not match: The names in x=... and y=... must match the CSV header. Check spelling and capitalization; runA and runa are different names.
• Putting the wrong value in the loop list: The loop items must be y-column names, not row numbers or data values. Keep the shared x-column in the x=... option.
• Using decimal commas without accounting for them: In this format, commas delimit fields, so decimal values should normally use a period, such as 2.2. A file that also uses commas as decimal marks needs a different, unambiguous delimiter and matching table settings.
• Expecting the loop to create legend entries automatically: \addplot draws a series, but a named legend entry requires \addlegendentry or another legend configuration.

If each column needs a different color or mark, PGFPlots can cycle styles automatically across successive \addplot commands. You can also set styles directly in the loop or configure the cycle list for the axis.