如何在KSH脚本中用Sed仅替换非嵌套/* */内的//注释？

Hey there, I see you're dealing with a tricky comment conversion issue when migrating your Pro C code from Windows to Linux. Your current sed script is replacing all // comments—even those nested inside /* */ blocks—which creates broken comment structures like /* /* some comments */ code; code; */ that throw off compilers. Let's fix this properly.

Why Your Current Sed Script Fails #

Sed operates line-by-line and can't track state (like whether you're inside a multi-line comment block). It has no way to distinguish between a // in regular code vs. one nested inside an existing /* */ comment, leading to those invalid nested comment messes. We need a tool that can track this state, and awk is perfect for the job.

The Awk Solution #

This awk script will only replace // comments that are not inside /* */ blocks, while handling multi-line comments correctly:

#!/usr/bin/awk -f

BEGIN {
    in_block_comment = 0  # Track if we're inside a /* */ comment
}

{
    current_line = $0
    remaining_line = current_line

    while (length(remaining_line) > 0) {
        if (!in_block_comment) {
            # Find positions of // and /* in the remaining line
            pos_double_slash = index(remaining_line, "//")
            pos_block_start = index(remaining_line, "/*")

            # Prioritize // if it comes before /*
            if (pos_double_slash > 0 && (pos_block_start == 0 || pos_double_slash < pos_block_start)) {
                # Replace // with /* ... */
                print substr(remaining_line, 1, pos_double_slash - 1) "/*" substr(remaining_line, pos_double_slash + 2) "*/"
                remaining_line = ""
            }
            # Handle starting a block comment
            else if (pos_block_start > 0) {
                print substr(remaining_line, 1, pos_block_start + 1)
                remaining_line = substr(remaining_line, pos_block_start + 2)
                in_block_comment = 1
            }
            # No comments left, print the rest
            else {
                print remaining_line
                remaining_line = ""
            }
        }
        else {
            # Look for the end of the block comment
            pos_block_end = index(remaining_line, "*/")
            if (pos_block_end > 0) {
                print substr(remaining_line, 1, pos_block_end + 1)
                remaining_line = substr(remaining_line, pos_block_end + 2)
                in_block_comment = 0
            }
            # No end found, print the whole line (still in block comment)
            else {
                print remaining_line
                remaining_line = ""
            }
        }
    }
}

How This Script Works #

• State Tracking: The in_block_comment variable keeps track of whether we're inside a /* */ comment (1 = yes, 0 = no).
• Line Processing: For each line, we process it in chunks:
  • When not in a block comment, we check for // first—if it appears before any /*, we replace it with /* ... */.
  • If we hit a /*, we mark ourselves as being in a block comment and continue processing the rest of the line.
  • When inside a block comment, we only look for */ to exit the block—any // inside is ignored entirely.

Using the Script #

1. Save the script above as convert_comments.awk.
2. Make it executable:

   chmod +x convert_comments.awk
   

3. Run it on your Pro C files (replace $path with your actual directory):

   for file in "$path"/*.pc; do
       ./convert_comments.awk "$file" > "$file.tmp" && mv "$file.tmp" "$file"
   done
   

   We use a temp file to avoid overwriting the original file before confirming the conversion works.

Key Notes #

• Handles multi-line /* */ comments correctly (even if they span multiple lines).
• Ignores // comments nested inside /* */ entirely—no broken nested comments.
• If your code has // inside string literals (like printf("//hello");), this script will incorrectly replace them. If that's a case you need to handle, we can tweak the script to check for string boundaries, but it adds complexity. Let me know if you need that!

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