如何用sed高效去除美元符号单侧空格？支持跨行$$场景

Let's tackle this step by step—your initial attempts were close, but greedy matching and line-by-line processing were tripping you up. Here's how to fix both inline $...$ formulas and multiline $$...$$ blocks correctly:

1. Fix Inline $...$ Formulas #

The problem with your first capture group command was that .* greedily matches everything, including spaces and other $ characters. Instead, use [^$]* to match only characters that aren't a dollar sign, which ensures we stop at the closing $.

Run this GNU sed command:

sed -E 's/\$\s*([^$]*)\s*\$/\$\1\$/g' input.txt

Breakdown: #

• \$\s*: Matches an opening $ followed by any number of spaces/tabs
• ([^$]*): Captures all characters except $ (so it stops at the closing delimiter)
• \s*\$: Matches any number of spaces/tabs before the closing $
• \$\1\$: Replaces with the opening $, captured content, and closing $ (no extra spaces)

Testing your example:

Input: Some text $ latex code $ some more text
Output: Some text $latex code$ some more text

This won't touch standalone $ characters like $ some because they don't have a matching closing $ in the same line.

2. Fix Multiline $$...$$ Blocks #

For block formulas that span multiple lines, sed's default line-by-line processing won't work. Use GNU sed's -z flag to read the entire file as a single stream, and adjust the pattern to target $$ delimiters:

sed -zE 's/\$\$\s*([^\$]*?)\s*\$\$/\$\$\1\$\$/g' input.txt

Breakdown: #

• -z: Treats the input as a null-terminated stream (so we can match across lines)
• \$\$\s*: Matches opening $$ followed by any whitespace
• ([^\$]*?): Non-greedy capture of all characters except $ (stops at the first closing $$)
• \s*\$\$: Matches whitespace before the closing $$

This works even if your block formula looks like:

$$
  \begin{align}
    E = mc^2
  \end{align}
$$

It will trim the spaces around the content to:

$$
\begin{align}
  E = mc^2
\end{align}
$$

3. Combine Both Fixes in One Command #

To handle both inline and block formulas at once, chain the two substitutions:

sed -zE 's/\$\$\s*([^\$]*?)\s*\$\$/\$\$\1\$\$/g; s/\$\s*([^$]*)\s*\$/\$\1\$/g' input.txt

Notes: #

• This uses GNU sed features (-E, -z). If you're on macOS, install GNU sed via Homebrew (brew install gnu-sed) and use gsed instead of sed.
• If your LaTeX code contains escaped \$ characters (literal dollar signs), you'll need to adjust the regex to ignore those—let me know if that's a case you need to handle!

内容的提问来源于stack exchange，提问作者physicophilic