diff --git a/cpdfdraw.ml b/cpdfdraw.ml index a98b27b..a18a18b 100644 --- a/cpdfdraw.ml +++ b/cpdfdraw.ml @@ -169,7 +169,6 @@ let process_specials pdf endpage filename bates batespad num page s = in Cpdfaddtext.process_text (res ()).time s pairs -(* FIXME cache (just for paragraph) *) let font_widths f fontsize = match f with | Pdftext.StandardFont (sf, encoding) -> @@ -191,9 +190,6 @@ let runs_of_utf8 s = let triples = option_map (Cpdfembed.get_char fontpack) codepoints in let collated = Cpdfembed.collate_runs triples in let font_widths fontnum font font_size = - (* Need to cache font widths here. TODO need to cache futher up too for - more speed. Check -typeset speed now we need widths. Or, make width - calculation optional? *) match Hashtbl.find_opt widthcache (fontnum, font_size) with | Some table -> table | None -> @@ -290,6 +286,24 @@ let structdata = ref [] (* TODO: Tagging in XObjects, move tag state into res () etc. *) +let rec remove_tfs prev = function + | [] -> [] + | Pdfops.Op_Tf (f, _)::t when f = prev -> remove_tfs prev t + | Pdfops.Op_Tf (f, s) as h::t -> h::remove_tfs f t + | h::t -> h::remove_tfs prev t + +let rec merge_adjacent_tjs ops = + let merge_tjs l = + Pdfops.Op_Tj (String.concat "" (map (function Pdfops.Op_Tj s -> s | _ -> assert false) l)) + in + match cleavewhile (function Pdfops.Op_Tj _ -> true | _ -> false) ops with + | [], h::t -> h::merge_adjacent_tjs t + | [], [] -> [] + | l, t -> merge_tjs l::merge_adjacent_tjs t + +let clean_up ops = + merge_adjacent_tjs (remove_tfs "" ops) + (* TODO: Use Uuseg for proper unicode segmentation. *) let format_paragraph indent j w s = let ss = String.split_on_char ' ' s in @@ -334,7 +348,7 @@ let format_paragraph indent j w s = end; done; allops =| rev (Pdfops.Op_T'::justify !ops); - flatten (rev !allops) + clean_up (flatten (rev !allops)) let current_eltinfo = null_hash ()