Freeform text profile field does not migrate properly from D6 to D7

This patch makes sure the data splits on newlines and commas, fixing the issue. This patch includes updated test files.

The patch looks good to me - thanks for adding the additional tests!

-mike

1. +++ b/core/modules/migrate_drupal/src/Plugin/migrate/source/d6/ProfileFieldValues.php
   @@ -53,7 +53,8 @@ public function prepareRow(Row $row) {
   +        // Explode by newline and comma
   

   Missing full stop.

2. +++ b/core/modules/migrate_drupal/src/Plugin/migrate/source/d6/ProfileFieldValues.php
   @@ -53,7 +53,8 @@ public function prepareRow(Row $row) {
   +        $row->setSourceProperty($profile_value['name'], preg_split("/(,)|((\r(?!\n))|((?<!\r)\n)|(\r\n))/", $profile_value['value']));
   

   This regex looks a bit much? How about:

   /[\r\n,]+/

   from: http://stackoverflow.com/a/3483314

Benjy,

That's a pretty little bit of RegEx that you found. Elegant solution.

I also beefed up the test a bit.

-mike

+++ b/core/modules/migrate_drupal/src/Tests/Dump/Drupal6User.php
@@ -450,7 +450,7 @@ public static function getData($table) {
+        array('fid' => 12, 'uid' => 2, 'value' => "AC/DC\n,,Eagles\r\nElton John,Lemonheads\r\n\r\nRolling Stones\rQueen"),

Do we need just a \n on it's own?

Sounds like a good idea. Updated patch attached.

Thanks,
-mike

I used that long regex so it would catch single \n and \r\n. Some operating systems still produce \n, some produce \r and most \r\n (some info here: http://stackoverflow.com/questions/13092837/does-n-r-behave-differently-...).

The original plan was to use PHP_EOL, but I dont know how to put that in a regex.
Just using \r\n might break in some cases, the patch in #3 searches for \r\n and after that for \r and \n, so it will work in all cases.

As mentioned on the call I think it might be worth moving some of the stuff in prepareRow() into a process plugin at some point but that shouldn't hold this issue up since we want people to keep testing their migrations and this fixes the issue.

jjpost, the regex we have now searches for 1 or more of the characters in the regex so it handles the \n and \r cases as well as \r\n etc. This is shown by the tests still passing.

I don't think PHP_EOL would have worked since that is picked up from the current environment and that might be different from when the data was created.

@benjy You are right, patch #8 works perfectly. Thanks!

Committed 5027a94 and pushed to 8.x. Thanks!