From 75173a030b1f3eb8a1e161ffd1732daa40a5e178 Mon Sep 17 00:00:00 2001 From: Nick White Date: Wed, 17 Apr 2019 17:51:34 +0100 Subject: Add basic dehyphenate tool --- dehyphenate/main.go | 63 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 63 insertions(+) create mode 100644 dehyphenate/main.go (limited to 'dehyphenate') diff --git a/dehyphenate/main.go b/dehyphenate/main.go new file mode 100644 index 0000000..4393c8f --- /dev/null +++ b/dehyphenate/main.go @@ -0,0 +1,63 @@ +package main + +import ( + "encoding/xml" + "flag" + "fmt" + "io/ioutil" + "log" + "os" + + "rescribe.xyz/go.git/lib/hocr" +) + +// BUGS: +// - loses all elements not captured in hocr structure such as html headings +// might be best to copy the header and footer separately and put the hocr in between, but would still need to ensure all elements are captured +// - loses any formatting; doesn't need to be identical, but e.g. linebreaks after elements would be handy +// - need to handle OcrChar + +func main() { + flag.Usage = func() { + fmt.Fprintf(os.Stderr, "Usage: dehyphenate hocrin hocrout\n") + fmt.Fprintf(os.Stderr, "Dehyphenates a hocr file.\n") + flag.PrintDefaults() + } + flag.Parse() + if flag.NArg() != 2 { + flag.Usage() + os.Exit(1) + } + + in, err := ioutil.ReadFile(flag.Arg(0)) + if err != nil { + log.Fatalf("Error reading %s: %v", flag.Arg(1), err) + } + h, err := hocr.Parse(in) + if err != nil { + log.Fatal(err) + } + + for i, l := range h.Lines { + w := l.Words[len(l.Words)-1] + if len(w.Chars) == 0 { + if len(w.Text) > 0 && w.Text[len(w.Text) - 1] == '-' { + h.Lines[i].Words[len(l.Words)-1].Text = w.Text[0:len(w.Text)-1] + h.Lines[i+1].Words[0].Text + h.Lines[i+1].Words[0].Text = "" + } + } else { + log.Printf("TODO: handle OcrChar") + } + } + + f, err := os.Create(flag.Arg(1)) + if err != nil { + log.Fatalf("Error creating file %s: %v", flag.Arg(1), err) + } + defer f.Close() + e := xml.NewEncoder(f) + err = e.Encode(h) + if err != nil { + log.Fatalf("Error encoding XML: %v", err) + } +} -- cgit v1.2.1-24-ge1ad