How to grep umlauts and other accented text characters via AppleScript

Question

I have a problem trying to execute shell scripts from apple script. I do a "grep", but as soon as it contains special characters it doesn't work as intended. (The script reads a list list ob subfolders in a directory and checks if any of the subfolders appear in a file.)

Here is my script:

set searchFile to "/tmp/output.txt"

set theCommand to "/usr/local/bin/pdftotext -enc UTF-8 some.pdf" & space & searchFile
do shell script theCommand

tell application "Finder"
    set companies to get name of folders of folder ("/path/" as POSIX file)
end tell

repeat with company in companies
    set theCommand to "grep -c " & quoted form of company & space & quoted form of searchFile

    try
        do shell script theCommand
        set CompanyName to company as string
        return CompanyName
    on error

    end try
end repeat

return false

The problem is e.g. with strings with umlauts. "theCommand" is somehow differently encoded that when I do it on the CLI directly.

$ grep -c 'Württemberg' '/tmp/output.txt' --> typed on command line
3
$ grep -c 'Württemberg' '/tmp/output.txt' --> copy & pasted from AppleScript
0
$ grep -c 'rttemberg' '/tmp/output.txt'   --> no umlauts, no problems
3

The "ü" from the first and the second line are different; a echo 'Württemberg' | openssl base64 shows this.

I tried several encoding tricks at different places, basically everything I could find or think of.

Does anyone have any idea? How can I check which encoding a string has?

Thanks in advance! Sebastian

Answer 1

Overview

This can work by escaping each character that has an accent in each company name before they are used in the grep command.

So, you'll need to escape each one of those characters (i.e. those which have an accent) with double backslashes (i.e. \\). For example:

• The ü in Württemberg will need to become \\ü
• The ö in Königsberg will need to become \\ö
• The ß in Einbahnstraße will need to become \\ß

---

Why is this necessary:

These accented characters, such as a u with diaeresis, are certainly getting encoded differently. Which type of encoding they receive is difficult to ascertain. My assumption is that the encoding pattern used begins with a backslash - hence why escaping those characters with backslashes fixes the issue. Consider the u with diaeresis in the previous link, it shows that for the C/C++ language the ü is encoded as \u00FC.

---

Solution

In the complete script below you'll notice the following:

1. set accentedChars to {"ü", "ö", "ß", "á", "ė"} has been added to hold a list of all characters that will need to be escaped. You'll need to explicitly state each one as there doesn't seem to be a way to infer whether the character has an accent.

2. Before assigning the grepcommand to theCommand variable we firstly escape the necessary characters via the line reading:

   set company to escapeChars(company, accentedChars)
   

   As you can see here we are passing two arguments to the escapeChars sub-routine, (i.e. the non-escaped company variable and the list of accented characters).

3. In the escapeChars sub-routine we iterate over each char in the accentedChars list and invoke the findAndReplace sub-routine. This will escape any instances of those characters with backslashes found in the company variable.

Complete script:

set searchFile to "/tmp/output.txt"
set accentedChars to {"ü", "ö", "ß", "á", "ė"}

set theCommand to "/usr/local/bin/pdftotext -enc UTF-8 some.pdf" & ¬
  space & searchFile
do shell script theCommand

tell application "Finder"
  set companies to get name of folders of folder ("/path/" as POSIX file)
end tell

repeat with company in companies
  set company to escapeChars(company, accentedChars)

  set theCommand to "grep -c " & quoted form of company & ¬
    space & quoted form of searchFile

  try
    do shell script theCommand
    set CompanyName to company as string
    return CompanyName
  on error

  end try
end repeat

return false

(**
 * Checks each character of a given word. If any characters of the word
 * match a character in the given list of characters they will be escapd.
 *
 * @param {text} searchWord - The word to check the characters of.
 * @param {text} charactersList - List of characters to be escaped.
 * @returns {text} The new text with the item(s) replaced.
 *)
on escapeChars(searchWord, charactersList)
  repeat with char in charactersList
    set searchWord to findAndReplace(char, ("\\" & char), searchWord)
  end repeat
  return searchWord
end escapeChars

(**
 * Replaces all occurances of findString with replaceString
 *
 * @param {text} findString - The text string to find.
 * @param {text} replaceString - The replacement text string.
 * @param {text} searchInString - Text string to search.
 * @returns {text} The new text with the item(s) replaced.
 *)
on findAndReplace(findString, replaceString, searchInString)
  set oldTIDs to text item delimiters of AppleScript
  set text item delimiters of AppleScript to findString
  set searchInString to text items of searchInString
  set text item delimiters of AppleScript to replaceString
  set searchInString to "" & searchInString
  set text item delimiters of AppleScript to oldTIDs
  return searchInString
end findAndReplace

---

Note about current counts:

Currently your grep pattern only reports the number of lines that the word was found on. Not how many instances of the word were found.

If you want the actual number of instances of the word then use the -o option with grep to output each occurrence. Then pipe that to wc with the -l option to count the number of lines. For example:

grep -o 'Württemberg' /tmp/output.txt | wc -l

and in your AppleScript that would be:

set theCommand to "grep -o " & quoted form of company & space & ¬
  quoted form of searchFile & "| wc -l"

Tip: If your want to remove the leading spaces in the count/number that gets logged then pipe it to sed to strip the spaces: For example via your script:

set theCommand to "grep -o " & quoted form of company & space & ¬
  quoted form of searchFile & "| wc -l | sed -e 's/ //g'"

and the equivalent via the command line:

grep -o 'Württemberg' /tmp/output.txt | wc -l | sed -e 's/ //g'