Sed expression to convert only certain lines starting with some example phrase

There's an example file with a regular text. In some places of the document there's a mix of the following lines:

| ![](/img/2016/12/020.jakis-tam-text1.png#medium) | ![](/img/2016/12/021.jakis-tam-text2.png#medium) | ![](/img/2016/12/022.jakis-tam-text3.png#medium) |
| ![](/img/2016/12/020.jakis-tam-text1.png#medium) | ![](/img/2016/12/021.jakis-tam-text2.png#medium) |

There's the following sed expression to convert the lines to the required form:

sed 's#\([^[]*.\)\([^\.]*.\([^\.]*\)[^)]*.\)#\1\3\2#g'

How to apply this sed expression to only those lines that start with | ![]( ?

Your pattern is not quite right: . when unescaped matches any char. Also, note you do not need to escape . chars inside bracket expressions.

I suggest the following command:

sed '/^|[[:space:]]!\[](/s#\(|[[:space:]]!\[\)\(]([^().]*\.\([^|.]*\)\.[^()]*)\)#\1\3\2#g'

See online demo.

Here, the POSIX BRE pattern matches

• /^|[[:space:]]!\[](/ - any line starting with |, whitespace, and then ![]( text
• s#\(|[[:space:]]!\[\)\(]([^().]*\.\([^|.]*\)\.[^()]*)\)#\1\3\2#g -
  • \(|[[:space:]]!\[\) - Group 1 (\1): |, a whitespace, ![ text
  • \(]([^().]*\. - Group 2 start: ](, then any zero or more chars other than (, ) and ., then a . (note it is escaped)
    • \([^|.]*\) - Group 3 (\3): any zero or more chars other than | and .
  • \.[^()]*)\) - (still Group 2): . char, then zero or more chars other than ( and ) and then a ) char.

The replacement is the concatenation of Group 1, 3 and 2.