6

I'm totally mystified why I can't print this range with ed when there are two ranges—see file2.tf file—; but I can print when there is only one range—see file1.tf file—and I can print with the gsed (GNU sed on macOS) command; but I can't print with ed. Please consider my shell session and clarify my misconceptions:

$ ed -s file1.tf <<<',n'
1   # some comment
2   module "hello_world" {
3     source = "./mydir"
4   }
5   # another comment
$ ed -s file1.tf <<<'/module.*world/,/}/p'
module "hello_world" {
  source = "./mydir"
}
$ ed -s file2.tf <<<',n'
1   # some comment
2   module "hello_world" {
3     source = "./mydir"
4   }
5   # another comment
6   # some comment
7   module "hello_again" {
8     source = "./anotherdir"
9   }
10  # another comment
$ ed -s file2.tf <<<'/module.*again/,/}/p'
?
$ gsed -n '/module.*again/,/}/p' file2.tf
module "hello_again" {
  source = "./anotherdir"
}

Update: The reverse direction works; but I'm not sure why:

$ ed -s file2.tf <<<'?module.*again?,?}?p'
module "hello_again" {
  source = "./anotherdir"
}

Update 2: The ?? approach doesn't actually work as intended (if there are three sections like in file3.tf file example) and see answer for an explanation.

$ ed -s file3.tf <<<,n
1   # some comment
2   module "hello_world" {
3     source = "./mydir"
4   }
5   # another comment
6   # some comment
7   module "hello_again" {
8     source = "./anotherdir"
9   }
10  # another comment
11  # some comment
12  module "hello_yet_again" {
13    source = "./yetanotherdir"
14  }
15  # another comment
$ ed -s file3.tf <<<'?module.*hello_again?,?}?p'
module "hello_again" {
  source = "./anotherdir"
}
# another comment
# some comment
module "hello_yet_again" {
  source = "./yetanotherdir"
}
Improve this question

1 Answer 1

Reset to default
10

The difference between ed and sed that causes your issue is how regular expression address ranges are handled.

In sed, the start line is first found, and the command is applied to all lines until the end line is found.

In ed, on the other hand, the start and stop line for the range is computed relative to the current line. The command is then applied to all lines in that range of lines.

In your case with /module.*again/, /}/ in ed, and with your second file, the current line is at the end of the file (since you just loaded it into the buffer). The start of the range is computed as line number 7, and the end line is line number 4 (the first line after the current line that matches the second expression). You can't have a range that runs backwards, so you get an error.

With ?module.*again?, ?}?, the search happens in reverse, so the start of the range is computed as line 7 as before (there is only a single line matching the expression module.*again), but now the end of the range is line 9, which is larger than 7, so you don't have the same issue.

Your solution is not to reverse the search by using ? in place of / as the regular expression delimiters, as that would fail to properly find the middle section of three sections (you only have two in your examples). Instead, first, move the cursor to the first line of the range, then apply the command from that line to the end of the section:

/module.*again/; /}/ p

This looks very similar to /module.*again/, /}/ p, but by using ; instead of ,, the end address is not calculated until the start address has been found. This is how sed handles ranges by default, as it has no other choice (without storing the complete document in memory).

The expression is essentially the same as a longer expression that explicitly moves the cursor to the start of the range, then applies the command from the now current line to the end of the section:

/module.*again/; ., /}/ p

The POSIX specification for ed says this about , vs. ;:

Addresses shall be separated from each other by a <comma> ( , ) or <semicolon> character ( ; ). In the case of a <semicolon> separator, the current line ( . ) shall be set to the first address, and only then will the second address be calculated. This feature can be used to determine the starting line for forwards and backwards searches.

Improve this answer
2
  • Excellent explanation! And thank you for your care to explain my mistake about the ?? backwards search; I added an example with 3 sections to illustrate that unintended result. Also, I applied the n command since you mentioned line numbers. Checking my understanding: Your solution is to realize that ed reads the full file into buffer and lands at the end; therefore approach by wrap to the target pattern begin address, then print from there to target pattern end address. Also, I'm glad to see your fosstodon.org/@kusalananda/111702532344757014 post! Commented Jan 5, 2024 at 18:36
  • Just a follow-up: The Semicolon ‘;’ section of the wolfram.schneider.org/bsd/7thEdManVol2/adv.ed/adv.ed.pdf tutorial has a detailed explanation of the difference between comma and semicolon for address ranges. Commented Jan 25 at 20:04

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.