Delete lines between 2 patterns, but only if it has a specific string inbetween

Question

Example, I have a file with the following contents:

eggs
bacon
cereal

eggs
bacon
cheese
cereal

Output:

eggs
bacon
cereal

In the example, I want to delete lines between the words eggs and cereal, but only if it contains cheese in between.

How could I do this using sed?

I have been asked to be more specific, so here is basically what I want to do, and I hope this makes it more clear:

WeaponData
{
    TextureData
    {
                "crosshair"
                {
                                "file"  "vgui/replay/thumbnails/"
                                "x"             "0"
                                "y"             "0"
                                "width"         "64"
                                "height"        "64"
                }
                "weapon"
                {
                                "file"          "sprites/bucket_bat_red"
                                "x"             "0"
                                "y"             "0"
                                "width"         "200"
                                "height"                "128"
                }
                "weapon_s"
                {       
                                "file"          "sprites/bucket_bat_blue"
                                "x"             "0"
                                "y"             "0"
                                "width"         "200"
                                "height"                "128"
                }

                "ammo"
                {
                                "file"          "sprites/a_icons1"
                                "x"             "55"
                                "height"    "15"
                }

                "crosshair"
                {
                    "file"      "sprites/crosshairs" <---
                    "x"     "32"
                    "y"     "32"
                    "width"     "32"
                    "height"    "32"
                }

                "autoaim"
                {
                    "file"      "sprites/crosshairs"
                    "x"     "0"
                    "y"     "48"
                    "width"     "24"
                    "height"    "24"
                }
    }
}

This is the command I tried:

sed '/"crosshair"/,/}/d' file.txt

That command just deletes from the very first "crosshair" to the very last }, regardless of whether it has the line "sprites/crosshairs" inbetween or not.

I only want to delete from "crosshair" to } if between the patterns, the string "sprites/crosshairs" is found. However there are other blocks of code on the text file containing "sprites/crosshairs".

Such as:

"autoaim"
{
        "file"      "sprites/crosshairs"
        "x"     "0"
        "y"     "48"
        "width"     "24"
        "height"    "24"
}

This cannot be removed. I DEEPLY apologize for not explaining this earlier, but I did not think it would be necessary.

don_crissti's suggestion works quite well which is sed '/crosshair/,/\}/{H;/\}/!d;s/.*//;x;/sprites\/crosshairs/d;s/.//;}' infile

The output is this, however:

              ...
              "autoaim"
              {
  }
}

"autoaim" is partially deleted, but not completely, but as you can see the block containing "sprites/crosshairs" is removed, as I wanted to do, but autoaim still cannot be modified

Answer 1

sed '/^[[:blank:]]*"crosshair"/,/}/{H;/}/!d;s/.*//;x;/sprites\/crosshairs/d;s/.//;}' infile

How it works:

sed '/^[[:blank:]]*"crosshair"/,/}/{         # in this range
H                                            # append each line to hold buffer
/}/!d                                        # delete it if not the end of range
s/.*//                                       # empty the pattern space
x                                            # exchanges buffers
/sprites\/crosshairs/d                       # delete pattern space if it matches
s/.//                                        # remove leading newline, autoprint
}' infile

This assumes lines matching ^[[:blank:]]*"crosshair" are always followed by a block of lines enclosed in braces, just like in your example.

Answer 2

if perl is okay, one can use paragraph mode (-00 option) and use multi-line regex match

$ cat ip.txt 
eggs
bacon
cereal

eggs
bacon
cheese
cereal

$ perl -00 -ne 'print if !/^eggs.*cheese.*cereal$/ms' ip.txt 
eggs
bacon
cereal