Are there some Latex style guides with part dedicating to labeling? I want to use some persistent labeling scheme, but don't know any (except for uncertain one in my mind).

What labeling scheme should I use and why? Some (but not all) of questions I want to find answer to:

  1. Should it be : to separate labeled environment from label itself (eg thm:label) and why?
  2. Should I use - or _ or something else as separation (eg eq:my_equations vs eq:my-equation vs eq:MyEquation)?
  3. How do I do nesting labeling? (eg equations in theorem environment)

I want to find some mature labeling schemes which is good for reading, autocompletion in editors and refactoring. I would be grateful for any references.

  • 1
    Welcome to TeX - LaTeX! There is a risk that what you ask for is pure opinion. Concrete variations would include things like which characters does LaTeX allow in labels? Which labelling schemes are provided automatically by systems such as RefTeX? Mar 10 '16 at 9:05
  • @AndrewSwann, thanks for welcoming. I consider my question as version of tex.stackexchange.com/questions/577/best-practices-references but with focus on labeling.
    – petRUShka
    Mar 10 '16 at 9:17
  • As long as the label only contains allowed characters, it doesn't matter how you name it. There is no advatage of using : over ;, just make the labels mnemonic.
    – Johannes_B
    Mar 10 '16 at 9:20
  • you only need to label things you reference and you only need to give them a name that you can remember long enough to enter \ref at the point where you need the reference, so no scheme is needed really. Mar 10 '16 at 9:24
  • 1
    I would avoid using colons in ref names because of this. Though if you don't use babel or cleveref it's not a problem. Mar 10 '16 at 15:20

1) From my personal experience, it is helpful to start with the category (i.e. thm: for theorems, eq: for equations, fig: for figures...) as you might have an equation and a figure (for example) which go together so you can call them the same and distinguish them by category.

2) For the name of the label I like to use underscore to separate words but it is only because I find very confusing not using separating characters.

3) I don't use nesting labelling in theorems but, when I have a system of equations, I use subequation environment (which labels as 1a, 1b...) and label the whole set of equations so when I reference it I get the number (1 in my example) and then I just add the letter manually. I do that because then I avoid having a lot of labels and it usually works well because I don't introduce changes in the ordering of those equations.

I hope this helps you!

  • 1
    One reason to use prefixes like thm: is that editors (e.g. Emacs+AUCTeX) recognize these and then can automatically offer completion to the correct types of labels only. Mar 10 '16 at 11:59

I use the following.

1) I agree with @FerranBrosaPlannella that it is a good idea to use prefixes for categories. I also think that the prefix should be separated from the name and : is kind of standard for this. I don't think that categories like Theorem, Proposition, Lemma, Observation,… should have distinct prefixes. They mean essentially the same and I sometimes change my mind whether something should be Proposition or Lemma, so I use thm: prefix for all of them. Similarly I use def: prefix for Definition, Notation, Convention, …

2) I also like when the word are separated and I use _ for this. I sometimes use - in labels for something else. E.g. there is something called “I-space”, and if there was a theorem about decomposition of I-spaces, I would label it thm:I-space_decomposition.

3) I use nested labeling for example when I have a theorem about equivalence between several conditions. The theorem has its label and also the conditions have labels, so they can be referenced individually from outside, but also locally in the proof. I separate the nested parts of the label by ., so such label may look like thm:some_equivalence.condition_1. Such labels are long, which is impractical when used locally, so I define a set of macros to work with these local labels. An example of usage follows.

\begin{theorem} \labelblock{thm:main}
    The following are equivatent.
        \item First condition. \loclabel{first}
        \item Conjunction of the following: \loclabelblock{second}
            \item First part. \loclabel{part_one}
            \item Second part. \loclabel{part_two}

        We can refer to item \locref{first} or subitem \locref{second.part_one}. To prove \locequiv{first}{second} we start with showing \locimpl{first}{second.part_one}.

Now we refer to Theorem \ref{thm:main} and to and item in a theorem: Theorem \itemref{thm:main}{second.part_two}. Now we refer to the item only: \ref{thm:main.second.part_two}.

The macro \labelblock does the same thing as \label, but additionally stores the label, so \loclabel and \locref can use it. \locref{first} expands to \ref{thm:main.first}. When multiple levels of nesting are needed (which is quite rare), \loclabelblock is used – it is a combination of \loclabel and \labelblock. The \locimple and \locequiv macros are just local versions of my macros for referencing to implications and equivalences between conditions in theorems.


I use type_chapter_detail style notation. The label is also identical to the source filename (if the label points to an item brought in from a file by \input or \includegraphics), so fig_theory_bandstructure would be trivial to pair up with the .pdf figure it points to. A check could easily be automated to ensure that all imported figures had both a \label and a \ref.

I tend to put tables in their own .tex files (\input{tab_theory_materials}\label{tab_theory_materials}), mainly so I can find my way around the source more easily, and compile just the table when I'm getting it right. This could easily be wrapped in a macro \input{#1}\label{#1} though I don't. This approach could be extended to complex equations.

I don't use any editor autocompletion features, so the benefits of fig:theory_bandgap style notation are lost to me.

  • I assume that by “if there is one” you mean “if there is just one label in the file”.
    – user87690
    Mar 10 '16 at 13:12
  • @user87690, if there is a file (e.g. for an equation there wouldn't be). I'll edit.
    – Chris H
    Mar 10 '16 at 14:58
  • But if a file contains more labels, then the file cannot be idenical to all of them.
    – user87690
    Mar 10 '16 at 15:15
  • @user87690 the figure/table file doesn't, even can't, contain any labels (well it could if it's a .tex file). It's the filename that matches the label where that file is included.
    – Chris H
    Mar 10 '16 at 15:50

One important consideration to note when assigning labels is spell checkers. Many spell checkers will pick these labels or their references as misspelled words.

Because of this, I tend to use just numbers. The numbers are often not recognized as misspelled words.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.