# indentation scope lines broken in algpseudocode

I am trying to provide a vertical line that shows the indentation in an algorithm. I'm following this answer, but when I add some math that is higher than the line height, the lines get broken. This is the result and I also provide the code. Can this issue be fixed?

\documentclass{article}

\usepackage{amsmath}                              % For \eqref
\usepackage{amssymb}                             % For mathematical symbols
\usepackage{amsbsy}                                % For bold symbol greek letters

\usepackage[noend]{algpseudocode}
\usepackage{algorithm}
\usepackage{algorithmicx}

\makeatletter
% This is the vertical rule that is inserted
\newcommand*{\algrule}[1][\algorithmicindent]{\makebox[#1][l]{\hspace*{.5em}\vrule height .75\baselineskip depth .25\baselineskip}}%

\newcount\ALG@printindent@tempcnta
\def\ALG@printindent{%
\ifnum \theALG@nested>0% is there anything to print
\ifx\ALG@text\ALG@x@notext% is this an end group without any text?
% do nothing
\addvspace{-3pt}% FUDGE for cases where no text is shown, to make the rules line up
\else
\unskip
% draw a rule for each indent level
\ALG@printindent@tempcnta=1
\loop
\algrule[\csname ALG@ind@\the\ALG@printindent@tempcnta\endcsname]%
\ifnum \ALG@printindent@tempcnta<\numexpr\theALG@nested+1\relax% can't do <=, so add one to RHS and use < instead
\repeat
\fi
\fi
}%
\usepackage{etoolbox}
% the following line injects our new indent handling code in place of the default spacing
\patchcmd{\ALG@doentity}{\noindent\hskip\ALG@tlm}{\ALG@printindent}{}{\errmessage{failed to patch}}
\makeatother
% end vertical rule patch for algorithmicx
\makeatother

% for bmatrix with alignment
\makeatletter
\renewcommand*\env@matrix[1][*\c@MaxMatrixCols c]{%
\hskip -\arraycolsep
\let\@ifnextchar\new@ifnextchar
\array{#1}}
\makeatother

\begin{document}

\begin{algorithm}
\caption{Arbitrary Algorithm}\label{IS2OSLS}
\begin{algorithmic}[1]
\Require A matrix $\mathbf{A}$ of size $m\times n$.
\Ensure Something.
\For{$i$ in $m$}
\State{$\mathbf{X} \gets \begin{bmatrix}[rrr] \boldsymbol x_i & \boldsymbol x_j & \ldots \end{bmatrix}^\intercal$} \Comment{get coordinates}

\For{$j$ in $n$}
\If{$i=j$}
\State Select a random action
\Else
\If{$i=j^3+1$}
\State Stay silent
\State $j \gets \begin{cases} a & b \\ c & d \end{cases}$
\Else
\State $i \gets \mathbf{J}^\intercal$
\State Break
\EndIf
\EndIf
\EndFor
\EndFor
\end{algorithmic}
\end{algorithm}

\end{document}


UPDATE

Based on your code, I give another example where you can see that the rules overlap the return statement.

\documentclass{article}

\usepackage{mathtools}
\usepackage{amssymb}
\usepackage{amsbsy}

\usepackage[noend]{algpseudocode}
\usepackage{algorithm}
\usepackage{algorithmicx}

\newcommand*{\algrule}[1][\algorithmicindent]{\hspace*{.5em}\vrule\vrule
width 0pt height .75\baselineskip depth .25\baselineskip\hspace*{\dimexpr#1-.5em}}

\makeatletter
\newcount\ALG@printindent@tempcnta
\def\ALG@printindent{%
\ifnum \theALG@nested>0% is there anything to print
\ifx\ALG@text\ALG@x@notext% is this an end group without any text?
% do nothing
\addvspace{-3pt}% FUDGE for cases where no text is shown, to make the rules line up
\else
\unskip
% draw a rule for each indent level
\ALG@printindent@tempcnta=1
\loop
\algrule[\csname ALG@ind@\the\ALG@printindent@tempcnta\endcsname]%
\ifnum \ALG@printindent@tempcnta<\numexpr\theALG@nested+1\relax% can't do <=, so add one to RHS and use < instead
\repeat
\fi
\fi
}%
\usepackage{etoolbox}
\patchcmd{\ALG@doentity}{\noindent\hskip\ALG@tlm}{\ALG@printindent}{}{\errmessage{failed to patch}}
\makeatother

\AtBeginEnvironment{algorithmic}{\lineskip0pt}

\newcommand*\Let[2]{\State #1 $\gets$ #2}

\begin{document}

\begin{algorithm}
\caption{Test 2
\label{alg:interact}}
\begin{algorithmic}[0]
\Require{$\mathcal{M} := \left\{ \mathcal{N}, \mathcal{E} \right\}$ and  $\mathcal{G}$}
\Statex
\Function{test}{$\mathcal{M}, \mathcal{G}$}
\Let{$\mathcal{S}$}{$\emptyset$}
\For{$e_i \in \mathcal{E}$}
\For{$g_i \in \mathcal{G}$}
\Let{$\mathcal{I}$}{$e_i \cap g_i$}
\If{$\mathcal{I} \neq \emptyset$}
\Let{$\left\{n_k \right\}$}{$\mathcal{I}$}
\Let{$\phantom{\left\{n_k \right\} }\mathllap{ \mathcal{S}}$}{$\mathcal{S} \bigcup_k \left\{ n_k, s_k\right\}$}
\Let{$\phantom{\left\{n_k \right\} }\mathllap{ \mathcal{N}}$}{$\mathcal{N} \bigcup_k \left\{n_k \right\}$}
\Let{$\phantom{\left\{n_k \right\} }\mathllap{\mathcal{T}}$}{subdivide($e_i, \mathcal{I}$)}
\Let{$\phantom{\left\{n_k \right\} }\mathllap{\mathcal{E}}$}{$\mathcal{E} \cup \mathcal{T}$}
\EndIf
\EndFor
\EndFor
\State \Return{$\mathcal{W}, \mathcal{N}, \mathcal{E}$}
\EndFunction
\end{algorithmic}
\end{algorithm}

\end{document}


Which gives

UPDATE 2

The new code is

\documentclass{article}

\usepackage{mathtools}
\usepackage{amssymb}
\usepackage{amsbsy}

\usepackage[noend]{algpseudocode}
\usepackage{algorithm}
\usepackage{algorithmicx}

\newcommand*{\algrule}[1][\algorithmicindent]{\hspace*{.5em}\vrule\vrule
width 0pt height \baselineskip depth .25\baselineskip\hspace*{\dimexpr#1-.5em}}

\makeatletter
\newcount\ALG@printindent@tempcnta
\def\ALG@printindent{%
\ifnum \theALG@nested>0% is there anything to print
\ifx\ALG@text\ALG@x@notext% is this an end group without any text?
% do nothing
\else
\unskip
% draw a rule for each indent level
\ALG@printindent@tempcnta=1
\loop
\algrule[\csname ALG@ind@\the\ALG@printindent@tempcnta\endcsname]%
\ifnum \ALG@printindent@tempcnta<\numexpr\theALG@nested+1\relax% can't do <=, so add one to RHS and use < instead
\repeat
\fi
\fi
}%
\usepackage{etoolbox}
\patchcmd{\ALG@doentity}{\noindent\hskip\ALG@tlm}{\ALG@printindent}{}{\errmessage{failed to patch}}
\makeatother

\AtBeginEnvironment{algorithmic}{\lineskip0pt}

\newcommand*\Let[2]{\State #1 $\gets$ #2}

\algtext*{EndWhile}% Remove "end while" text
\algtext*{EndFor}% Remove "end for" text
\algtext*{EndIf}% Remove "end if" text
\algdef{SE}[DOWHILE]{Do}{doWhile}{\algorithmicdo}[1]{\algorithmicwhile\ #1}%

% increase line spacing for algorithmicx
\usepackage{etoolbox}
\makeatletter
\expandafter\patchcmd\csname\string\algorithmic\endcsname{\itemsep\z@}
{\itemsep=0.2ex plus2pt}{}{}
\makeatother

\begin{document}

\showoutput

\begin{algorithm}
\caption{Test 2
\label{alg:interact}}
\begin{algorithmic}[0]
\Require{$\mathcal{M} := \left\{ \mathcal{N}, \mathcal{E} \right\}$ and  $\mathcal{G}$}
\Statex
\Function{test}{$\mathcal{M}, \mathcal{G}$}
\Let{$\mathcal{S}$}{$\emptyset$}
\For{$e_i \in \mathcal{E}$}
\For{$g_i \in \mathcal{G}$}
\Let{$\mathcal{I}$}{$e_i \cap g_i$}
\If{$\mathcal{I} \neq \emptyset$}
\Let{$\left\{n_k \right\}$}{$\mathcal{I}$}
\Let{$\phantom{\left\{n_k \right\} }\mathllap{ \mathcal{S}}$}{$\mathcal{S} \bigcup_k \left\{ n_k, s_k\right\}$}
\Let{$\phantom{\left\{n_k \right\} }\mathllap{ \mathcal{N}}$}{$\mathcal{N} \bigcup_k \left\{n_k \right\}$}
\Let{$\phantom{\left\{n_k \right\} }\mathllap{\mathcal{T}}$}{subdivide($e_i, \mathcal{I}$)}
\Let{$\phantom{\left\{n_k \right\} }\mathllap{\mathcal{E}}$}{$\mathcal{E} \cup \mathcal{T}$}
\EndIf
\EndFor
\EndFor
\State \Return{$\mathcal{W}, \mathcal{N}, \mathcal{E}$}
\EndFunction
\end{algorithmic}
\end{algorithm}

\end{document}


which gives

• Rules with a fixed height and depth are added, they don't adjust to the current line dimensions. I'm not sure it's possible to achieve this with such an approach. – egreg Jan 31 '17 at 11:50
• – Andrew Swann Jan 31 '17 at 12:03

Here is a solution building on the posted code. A completely different approach using tikz is to be found in Problems with vertical lines in algorithmicx.

To modify the current code you need to use rules that do not have a fixed height and depth. One way to achieve that is simply to write \vrule. However, you should ensure minimum sizes and need to make sure that the resulting lines do not get moved apart by the \lineskip mechanism. The code below covers all of your examples.

\documentclass{article}

\usepackage{mathtools}
\usepackage{amssymb}
\usepackage{amsbsy}

\usepackage[noend]{algpseudocode}
\usepackage{algorithm}
\usepackage{algorithmicx}

\usepackage{etoolbox}

\newcommand{\algruledefaultfactor}{.75}
\newcommand{\algstrut}[1][\algruledefaultfactor]{\vrule width 0pt
depth .25\baselineskip height #1\baselineskip\relax}
\newcommand*{\algrule}[1][\algorithmicindent]{\hspace*{.5em}\vrule\algstrut
\hspace*{\dimexpr#1-.5em}}

\makeatletter
\newcount\ALG@printindent@tempcnta
\def\ALG@printindent{%
\ifnum \theALG@nested>0% is there anything to print
\ifx\ALG@text\ALG@x@notext% is this an end group without any text?
% do nothing
\else
\unskip
% draw a rule for each indent level
\ALG@printindent@tempcnta=1
\loop
\algrule[\csname ALG@ind@\the\ALG@printindent@tempcnta\endcsname]%
\ifnum \ALG@printindent@tempcnta<\numexpr\theALG@nested+1\relax% can't do <=, so add one to RHS and use < instead
\repeat
\fi
\fi
}%

\patchcmd{\ALG@doentity}{\noindent\hskip\ALG@tlm}{\ALG@printindent}{}{\errmessage{failed to patch}}

\AtBeginEnvironment{algorithmic}{\lineskip0pt}

\newcommand*\Let[2]{\State #1 $\gets$ #2}
\newcommand*\Stateh{\State \algstrut[1]}

\begin{document}

\begin{algorithm}
\caption{Arbitrary Algorithm}\label{IS2OSLS}
\begin{algorithmic}[1]
\Require A matrix $\mathbf{A}$ of size $m\times n$.
\Ensure Something.
\For{$i$ in $m$}
\State{$\mathbf{X} \gets \begin{bmatrix*}[r] \boldsymbol x_i & \boldsymbol x_j & \ldots \end{bmatrix*}^\intercal$} \Comment{get coordinates}
\For{$j$ in $n$}
\If{$i=j$}
\State Select a random action
\Else
\If{$i=j^3+1$}
\State Stay silent
\State $j \gets \begin{cases} a & b \\ c & d \end{cases}$
\Else
\State $i \gets \mathbf{J}^\intercal$
\State Break
\EndIf
\EndIf
\EndFor
\EndFor
\end{algorithmic}
\end{algorithm}

\begin{algorithm}
\caption{Test 2
\label{alg:interact}}
\begin{algorithmic}[0]
\Require{$\mathcal{M} := \left\{ \mathcal{N}, \mathcal{E} \right\}$ and  $\mathcal{G}$}
\Function{test}{$\mathcal{M}, \mathcal{G}$}
\Let{$\mathcal{S}$}{$\emptyset$}
\For{$e_i \in \mathcal{E}$}
\For{$g_i \in \mathcal{G}$}
\Let{$\mathcal{I}$}{$e_i \cap g_i$}
\If{$\mathcal{I} \neq \emptyset$}
\Let{$\left\{n_k \right\}$}{$\mathcal{I}$}
\Let{$\hphantom{\left\{n_k \right\} }\mathllap{ \mathcal{S}}$}{$\mathcal{S} \bigcup_k \left\{ n_k, s_k\right\}$}
\Let{$\hphantom{\left\{n_k \right\} }\mathllap{ \mathcal{N}}$}{$\mathcal{N} \bigcup_k \left\{n_k \right\}$}
\Let{$\hphantom{\left\{n_k \right\} }\mathllap{\mathcal{T}}$}{subdivide($e_i, \mathcal{I}$)}
\Let{$\hphantom{\left\{n_k \right\} }\mathllap{\mathcal{E}}$}{$\mathcal{E} \cup \mathcal{T}$}
\EndIf
\EndFor
\EndFor
\Stateh \Return{$\mathcal{W}, \mathcal{N}, \mathcal{E}$}
\EndFunction
\end{algorithmic}
\end{algorithm}

\end{document}


I have rewritten the \algrule command, unpacking it from a box, so an isolated \vrule can fill the height and depth of the current line. A rule of zero width is then drawn to force that height and depth to be reasonable on standard lines. This rule is encapsulated in a newcommand \algstrut. Finally the algorithmic environment is patched to have \lineskip set to 0pt.

If you have two consecutive high lines these may now touch, but then you can fix this by adding a zero width rule of suitable height and depth to your text in these one-off cases. This can be done using \algstrut[1] or some large factor on the given line.

It also turns out to be necessary to have a taller strut on \State lines, so I have provided a new command \Stateh that includes this.

Incidentally I have removed the code for the bmatrix alignment, as the mathtools package now provides bmatrix* which has an alignment option. Also I have use \hphantom instead of \phantom in your example, as you are only interested in adjust the horizontal spacing.

• I tried your code and it almost work with my real algorithm. There are two issues, first the rules overlap with a return statement. I guess this could be solved by your example on adding a zero width rule? Could you provide an example on how to do this? In another algorithm I still found a couple of places where the lines are discontinuous. – aaragon Jan 31 '17 at 13:39
• I have updated the question showing the example where the rule overlap with the return statement. – aaragon Jan 31 '17 at 13:49
• Hi again, there's something else. I also had a bit of code to increase the spacing of within the algorithm. If I remove this code then your solution is perfect, but if I put it I still get the broken line in the return statement. Do you think that your answer can also consider this case? I modified my update 2 to show you this. – aaragon Jan 31 '17 at 14:46
• But on my end it doesn't work still. I get even a greater separation before the return statement. Are you sure it works on your end? What is missing? – aaragon Jan 31 '17 at 18:03
• I have now rearranged the answer covering all your examples. You can now just use \algstrut[1] instead of the previous \alghigher{....} and I have provide \Stateh that incoporates this. Yes, this is all a hack, as is the code you started with. Ultimately one would want to rewrite and restructure the algorithmicx code, so such hacking is not needed. – Andrew Swann Feb 1 '17 at 8:21