Regexp_replace PART of unicode (emoji) in Scala-Spark dataframe - regex

Question

I am trying to use spark regexp_replace to replace all emojis whose unicode starts with
\uD83D and replace just that part of the unicode with \uD83D, but I've had no luck.
Here is an example:
I want to take all instances of "😂" (which in Scala is \uD83D\uDE02) and replace those with " 😂". That's easy enough with one emoji and works with this code:
.select(functions.regexp_replace($"text2", "[(\uD83D\uDE02)]", " \uD83D\uDE02").as("split2"))
With the above code, if I have a string like this "😂😂😂😂" in the text2 column, it will turn it into " 😂 😂 😂 😂", which I can then easily split by space.
I want to apply this to ALL emojis that start with \uD83D so I would assume doing this should work, but it doesn't:
.select(functions.regexp_replace($"text2", "[(\uD83D)]", " \uD83D").as("split2"))
This does not affect the data in any way. Even the following does not affect the data in any way, with or without the parenthesis and/or brackets:
.select(functions.regexp_replace($"text2", "[(u)]", " \uD83D").as("split2"))
If I could replace just the first 6 characters of those unicode strings, the following in the text2 column: "😱😂😂😱" would result in " 😱 😂 😂 😱", which is exactly what I need.
Thanks for your help!

Answer 1

You can use this regex \\B\uD83D.{1} and replace with the captured group $1 and space. Then trim to get rid of the last space and split:
val df = Seq(
("😂😂😂😂"),
("😱😂😂😱")
).toDF("text2")
df.select(
split(
trim(regexp_replace($"text2", "(\\B\uD83D.{1})", "$1 ")),
" "
).as("split2")
).show
//+-----------------+
//| split2 |
//+-----------------+
//|[😂, 😂, 😂, 😂]|
//|[😱, 😂, 😂, 😱]|
//+-----------------+