PDFs generated out of Google Docs create CID Widths with float unlike most cases

where integer is used as widths.

Author: sambitdash

src/PDFontMetrics.jl

@@ -179,15 +179,15 @@ end
 get_font_widths(basefonts::CosName) = read_afm(convert(CDTextString, basefonts))
 
 function get_cid_font_widths(cosDoc::CosDoc, font::IDDRef{CosDict})
-    m = IntervalTree{UInt16, Int}()
+    m = IntervalTree{UInt16, Float32}()
     encoding = cosDocGetObject(cosDoc, font, cn"Encoding")
     desc = cosDocGetObject(cosDoc, font, cn"DescendantFonts") |> get
     w = cosDocGetObject(cosDoc, desc[1], cn"W")
     dw = cosDocGetObject(cosDoc, desc[1], cn"DW")
     # If widths are not specified or the font encoding is not Identity-H
     # widths cannot be extracted.
     if w === CosNull || encoding != cn"Identity-H"
-        return (dw === CosNull) ? CIDWidth() : CIDWidth(get(dw))
+        return (dw === CosNull) ? CIDWidth() : CIDWidth(Float32(get(dw)))
     end
     w = get(w)
     next = iterate(w)
@@ -201,16 +201,16 @@ function get_cid_font_widths(cosDoc::CosDoc, font::IDDRef{CosDict})
         if ecid isa Vector
             for wdo in ecid
                 width = get(wdo)
-                m[Interval(UInt16(ccid), UInt16(ccid))] = width
+                m[Interval(UInt16(ccid), UInt16(ccid))] = Float32(width)
                 ccid += 1
             end
         else
             (width, state) = iterate(w, state)
-            m[Interval(UInt16(bcid), UInt16(ecid))] = get(width)
+            m[Interval(UInt16(bcid), UInt16(ecid))] = Float32(get(width))
         end
         next = iterate(w, state)
     end
-    return (dw === CosNull) ? CIDWidth(m) : CIDWidth(m, get(dw))
+    return (dw === CosNull) ? CIDWidth(m) : CIDWidth(m, Float32(get(dw)))
 end
 
 get_character_width(n::CosName, afm::AdobeFontMetrics) =

src/PDFonts.jl

@@ -389,14 +389,13 @@ function read_cmap(stm::IO)
 end
 
 struct CIDWidth
-    imap::IntervalTree{UInt16, Int}
-    dw::Int
-    CIDWidth(m::IntervalTree{UInt16, Int}, tdw::Int) = new(m, tdw)
+    imap::IntervalTree{UInt16, Float32}
+    dw::Float32
 end
 
-CIDWidth(m::IntervalTree{UInt16, Int}) = CIDWidth(m, 1000)
-CIDWidth(tdw::Int) = CIDWidth(IntervalTree{UInt16, Int}(), tdw)
-CIDWidth() = CIDWidth(1000)
+CIDWidth(m::IntervalTree{UInt16, Float32}) = CIDWidth(m, 1000f0)
+CIDWidth(tdw::Float32) = CIDWidth(IntervalTree{UInt16, Float32}(), tdw)
+CIDWidth() = CIDWidth(1000f0)
 
 mutable struct PDFont
     doc::PDDoc

test/runtests.jl

@@ -13,7 +13,7 @@ using PDFIO.Common: read_pkcs12
 
 include("debugIO.jl")
 
-pdftest_ver  = "0.0.6"
+pdftest_ver  = "0.0.7"
 pdftest_link = "https://github.com/sambitdash/PDFTest/archive/v"*pdftest_ver
 
 zipfile = "pdftest-"*pdftest_ver
@@ -364,6 +364,24 @@ end
         end
     end
 
+    @testset "Floating point CIDWidth" begin
+        @test begin
+            filename="sample-google-doc.pdf"
+            DEBUG && println(filename)
+            resfile, template, filename = local_testfiles(filename)
+            doc = pdDocOpen(filename)
+            io = util_open(resfile, "w")
+            try
+                extract_text(io, doc)
+            finally
+                util_close(io)
+                pdDocClose(doc)
+            end
+            @test files_equal(resfile, template)
+            length(utilPrintOpenFiles()) == 0
+        end
+    end
+
     @testset "Non-standard CMap" begin
         @test begin
             filename="16-969_o7jp.pdf"