Specify encoding when loading type1 font

[unhammer]

c.f. #9 this lets us do things like

main = do
   …
    Right azsubscriptStructure <- readType1FontWithEncoding (OtherEncoding subscriptEncoding) 
                                                           "./azsubscripts.pfb" "./azsubscripts.afm"
    run (standardDocInfo { author = "me", compressed = False }) docRect $ do
        azsubscriptFont <- mkType1Font azsubscriptStructure
        makeDocument azsubscriptFont …

subscriptEncoding :: M.Map String Char
subscriptEncoding = M.fromList
  [ ("uni2080",'₀')
  , ("uni2081",'₁')
  , ("uni2082",'₂')
  , ("uni2083",'₃')
  , ("uni2084",'₄')
  , ("uni2085",'₅')
  , ("uni2086",'₆')
  , ("uni2087",'₇')
  , ("uni2088",'₈')
  , ("uni2089",'₉')
  ]

I use these helpers to pick fonts:

         txtWithFont fontPicker textThatMightHaveSubscripts
         forceNewLine
  where
    fontPicker c
        | c `elem` subscriptEncoding = Normal azsubscriptFont
        | otherwise              = Normal timesRoman

txtWithFont :: (Char -> MyParaStyles) -> Text -> Para MyParaStyles ()
txtWithFont pickFont text = for_ broken styleChar
  where
    broken = groupWithKey pickFont (T.unpack text)
    styleChar :: (MyParaStyles, String) -> Para MyParaStyles ()
    styleChar (style, substring) = do 
                        setStyle style
                        txt (T.pack substring)

groupWithKey :: Eq key => (elt -> key) -> [elt] -> [(key, [elt])]
groupWithKey classify lst = fmap decorate lst |> Data.List.Extra.groupOnKey fst |> map (second (map snd))
  where
    decorate elt = (classify elt, elt)

---

Note: there are some functions like getEncoding that I don't understand why are in IO, but I haven't changed them (perhaps the idea was to keep the api general enough to add encodings that need IO at a later point?).

And like I said in the comments of #9 I have no idea what the best API here is, but at least this PR exposes the functions necessary to get something working, without exposing that much of the Encodings module.

[unhammer]

However, there's some issue with the embedded fonts, they show fine in emacs and zathura, but Atril/Firefox/Chrome don't see the symbols.

If I do

$ gs -o output.pdf -sDEVICE=pdfwrite -dEmbedAllFonts=true fromHPDF.pdf 
The following warnings were encountered at least once while processing this file:
        A FontDescriptor has a missing or bad /FontName

   **** This file had errors that were repaired or ignored.
   **** The file was produced by:
   **** >>>>  <<<<
   **** Please notify the author of the software that produced this
   **** file that it does not conform to Adobe's published PDF
   **** specification.

then the output.pdf is fixed (but gs also outputs PDF-1.7 instead of 1.5 and changes the font from type 1 to type 1C according to pdffonts), but I don't understand where the error was – the font itself or something HPDF did.

---

EDIT: I tried just overwriting glyphs of letters a-j in an existing font encoded as AdobeStandard according to fontforge, and supplying the encoding

[("a",'₀'),("b",'₁'),("c",'₂'),("d",'₃'),("e",'₄'),("f",'₅'),("g",'₆'),("h",'₇'),("i",'₈'),("j",'₉')]

to HPDF, and then it works fine. Basically, it seems like if Encoding is AdobeStandard in fontforge, then firefox etc. understand it, but not if it's "Glyph Order" or "Symbol" or anything like that. So it seems there's something about the way HPDF inserts the font definition that leads pdf.js to assume a certain order of glyphs in the pfb (while for some reason zathura was able to read it), or maybe it can't find the ordering, I don't know. I'm guessing the error is in HPDF and not fontforge though.

In any case, it's good enough for me now, hopefully someone else can pick up the baton wrt. that glyph order; I've shaved enough of yak for one weekend :)

[hsyl20]

Thanks for your contribution! It looks good to me. Thanks for the documentation.

Note: there are some functions like getEncoding that I don't understand why are in IO, but I haven't changed them (perhaps the idea was to keep the api general enough to add encodings that need IO at a later point?).

Sorry but I have no idea. I'm not the package author, I'm not using it, and I'm barely maintaining it. :-)

[hsyl20]

Please tell me if/when you want a release.