如何在Bash中正确去除多行Markdown注释？

云上老周

2026-5-26

Fixing Markdown Comment Removal (Including Embedded & Multi-Line Comments)

Let's get this sorted out—your existing sed commands fall short either because they can't handle multi-line comments or they greedily delete too much content. Here are reliable solutions:

Why Your Previous Commands Failed

First sed command: sed -e 's///g' $1
Sed works line-by-line by default, so it can't properly match comments that span multiple lines. Also, non-greedy matching (.*?) isn't supported in standard sed, making this ineffective for complex cases.
Second command: cat $1 | tr '\n' '\r' | sed -e 's///g' | tr '\r' '\n'
The .* in sed is greedy—it matches from the very first  in the entire file, wiping out all content in between (including non-comment text). That's why you saw the unwanted mass deletion.

Working Solutions

1. Perl (Cross-Platform, Most Reliable)

Perl handles multi-line input and context-aware matching seamlessly. Run this command:

perl -0777 -pe 's/<!--(?:[^-]|-[^-]|--[^>])*-->//g' your_input_file.txt

-0777: Reads the entire file as a single block instead of line-by-line, perfect for multi-line comments.
The regex  precisely matches comments without getting tripped up by embedded <!-- sequences:
- [^-]: Any character that's not a hyphen
- -[^-]: A hyphen followed by something that's not a hyphen
- --[^>]: Two hyphens followed by something that's not a closing >

2. GNU Sed (Linux-Specific)

If you're using GNU Sed (common on Linux systems), you can use its null-separator mode to handle multi-line content:

sed -z 's/<!--\([^-]\|-[^-]\|--[^>]\)*-->//g' your_input_file.txt

-z: Treats the input as a stream separated by null characters, effectively reading the whole file at once.
The regex works the same way as the Perl version, ensuring we only match complete comment blocks.