A client has sent in a file where the text "These fine ffine ften ffton flen fflen" is cot correctly extracted. The text in the PDF is represented by:

[(\037)7(e)8(s)-3(e \036)13(n)7(e \035n)7(e \034)10(en \033)10(o)10(n \032)4(en \031)4(en)-25( )]TJ

The glyphs Th, fi, ff, ft, fl, fft, ffl are referenced by single glyph id's. The cmap file that maps the glyph id to a unicode value is is actually malformed. The cmap file look like this:

<19> <00660066006C>
<1A> <0066006C>
<1B> <006600660074>
<1C> <00660074>
<1D> <006600660069>
<1E> <00660069>
<1F> <00540068>
<20> <0020>
<65> <0065>
<6E> <006E>
<6F> <006F>

According to the specification for cmap <###> represents only one unicode value and as a result our parser will only read the first unicode value ignoring the rest. According the spec, a cid -> more then one unicode value should start off with a bracket when defining the unicode values, for example.

<19> [<00660066006C>]

I guess long story short I need to do further work to our cmap parser to correctly interpret range format. The document in question was encoded with Adobe cs7 which means we'll likely see more file like this as time goes on.