British Pound Sign £ causing PG::CharacterNotInRepertoire: ERROR: invalid byte sequence for encoding “UTF8”: 0xa3 - ruby

Question

When collecting information containing the British Pound Sign '£' from external sources such as my bank, via csv file, and posting to postgres using ActiveRecord, I get the error:
PG::CharacterNotInRepertoire: ERROR: invalid byte sequence for encoding “UTF8”: 0xa3
The 0xa3 is the hex code for a £ sign. The perceived wisdom is to clearly specify UTF-8 on the string whilst replacing invalid byte sequences..
string.encode('UTF-8', {:invalid => :replace, :undef => :replace, :replace => '?'})
This stops the error, but is a lossy fix as the '£' is converted into a '?'
UTF-8 is able to handle the '£' sign, so what can be done to fix the invalid byte sequence and persist the '£' sign?

Answer 1

I'm answering my own question thanks to Michael Fuhr who explained the UTF-8 byte sequence for the pound sign is 0xc2 0xa3. So, all you have to do is find each occurrence of 0xa3 (163) and place 0xc2 (194) in front of it...
array_bytes = string.bytes
new_pound_ptr = 0
# Look for £ sign
pound_ptr = array_bytes.index(163)
while !pound_ptr.nil?
pound_ptr+= new_pound_ptr # new_pound_ptr is set at end of block
# The following statement finds incorrectly sequenced £ sign...
if (pound_ptr == 0) || (array_bytes[pound_ptr-1] != 194)
array_bytes.insert(pound_ptr,194)
pound_ptr+= 1
end
new_pound_ptr = pound_ptr
# Search remainder of array for pound sign
pound_ptr = array_bytes[(new_pound_ptr+1)..-1].index(163)
end
end
# Convert bytes to 8-bit unsigned char, and UTF-8
string = array_bytes.pack('C*').force_encoding('UTF-8') unless new_pound_ptr == 0
# Can now write string to model without out-of-sequence error..
hash["description"] = string
Model.create!(hash)
I've had so much help on this stackoverflow forum, I hope I have helped somebody else.