I have found no evidence that Tika is the problem. I have found reason to suspect that Ruby 1.9.3.'s CSV is acting funny. This is my work-around for Ruby 1.9.3, maybe it will be useful to someone besides me.
class TikaCSV
def self.parse(s)
s.split(/\n(?="[^"])/).reduce([]) { |a, x| a += CSV.parse(x) }
end
end
I also wrote it up here:
http://djwonk.tumblr.com/post/34370338490/visions-of-comma-separated-values
