From fd8bd1da18b11794a810f881441b69e99a7ed3e2 Mon Sep 17 00:00:00 2001
From: Rob Pearce <rob@nethack.net>
Date: Sat, 23 Jul 2022 12:05:21 +1000
Subject: [PATCH] Improve word summary when doing OCR.

---
 scan | 55 +++++++++++++++++++++++++++++++++++++++----------------
 1 file changed, 39 insertions(+), 16 deletions(-)

diff --git a/scan b/scan
index 40d9f68..9396281 100755
--- a/scan
+++ b/scan
@@ -115,7 +115,7 @@ function usage() {
 
 OCRPROG_DEF="/opt/homebrew/bin/ocrmypdf"
 OCRPROG="$OCRPROG_DEF"
-P2TPROG="pdftotext"
+P2TPROG="/opt/homebrew/bin/pdftotext"
 DUPLEXOPTS=""
 MODE="scan"
 DIR=${PDFDIR:-"~/Documents"}
@@ -131,7 +131,7 @@ TEST=0
 FINYEARTAGS=""
 VERBOSE=0
 
-ALLARGS="$@"
+ALLARGS=""
 if [[ -e $RCFILE && $ALLARGS != *-h* ]]; then
     notify "processing $RCFILE"
 
@@ -149,8 +149,10 @@ if [[ -e $RCFILE && $ALLARGS != *-h* ]]; then
     ok
 fi
 
-validargs="cdf:hlmMops:Ttv"
-while getopts "$validargs" i $ALLARGS; do
+function processarg() {
+    local i=$1
+    local arg=$2
+
     case "$i" in
         v)
             VERBOSE=1;
@@ -182,16 +184,17 @@ while getopts "$validargs" i $ALLARGS; do
             ;;
         o)
             OCR=1;
+            info "OCR mode enabled"
             ;;
         O)
-            OCRPROG="$OPTARG"
+            OCRPROG="$arg"
             ;;
         p)
             PREVIEW=1;
             info "preview enabled"
             ;;
         s)
-            SCANNEROPTS="-scanner \"$OPTARG\"";
+            SCANNEROPTS="-scanner \"$arg\"";
             ;;
         T)
             info "temp mode"
@@ -201,21 +204,32 @@ while getopts "$validargs" i $ALLARGS; do
             info "test mode"
             TEST=1;
             ;;
-        -f)
-            info "tag ^b$OPTARG^p will use financial year paths"
-            FINYEARTAGS="$FINYEARTAGS $OPTARG"
+        f)
+            info "tag ^b$arg^p will use financial year paths"
+            FINYEARTAGS="$FINYEARTAGS $arg"
             ;;
         *)
             error "invalid argument:  $i";
             usage;
+            exit 1
             ;;
     esac
+
+}
+
+validargs="cdf:hlmMops:Ttv"
+while getopts "$validargs" i $ALLARGS; do
+    processarg "$i" "$OPTARG"
+done
+unset OPTIND
+
+while getopts "$validargs" i $*; do
+    processarg "$i" "$OPTARG"
 done
 shift $((OPTIND - 1))
 
 if [[ $OCR -eq 1 ]]; then
     if [[ -x $OCRPROG ]]; then
-        info "OCR mode enabled"
         OCR=1;
     else
         wtext="OCR requested but ^b$OCRPROG^p not found."
@@ -329,7 +343,7 @@ while [[ $finished -eq 0 ]]; do
     if [[ $VERBOSE -eq 1 ]]; then
         info "will run:  ${SCANLINE} -verbose -dir \"${DIR}\" -name \"${FILENAME}\" \"$DUPLEXOPTS\" \"$SCANNEROPTS\" $TAGS 2>&1"
     fi
-    notify "Scanning..."
+    notify "Scanning"
     OUTPUT=$( ${SCANLINE} -verbose -dir "${DIR}" -name "${FILENAME}" "$DUPLEXOPTS" "$SCANNEROPTS" $TAGS 2>&1)
     rv=$?
 
@@ -398,7 +412,7 @@ fi
 
 PREVIEWFILE=""
 if [[ $PDFFILE == *" "* || $PDFFILE == *$'\n'* ]]; then
-    itext="Scanned $NUMPAGES page(s) to "
+    itext="Scanned ^b$NUMPAGES^p page(s) to "
     count=1
     for x in $PDFFILE; do
         if [[ $count -eq 1 ]]; then
@@ -416,19 +430,28 @@ else
 fi
 
 if [[ $OCR -eq 1 ]]; then
+    notify "Running OCR"
     OCROPTS="--output-type pdfa -r -d -q --author scan.sh --keywords '$TAGS'"
 
     $OCRPROG $OCROPTS "${PREVIEWFILE}" "${PREVIEWFILE}" >/dev/null 2>&1
     if [[ $? -eq 0 ]]; then
+        ok
         if [[ -x $P2TPROG ]]; then
-            words=$($P2TPROG "${PREVIEWFILE}" /dev/stdout | wc -c)
+            allwords=$($P2TPROG "${PREVIEWFILE}" /dev/stdout)
+            words=$(echo "$allwords" | wc -w | bc)
+            mostfreq=$(echo "$allwords" | tr ' ' '\n' 2>/dev/null | tr -dc 'A-Za-z\n' 2>/dev/null | awk NF | egrep '[a-z]' | sort | uniq -c | sort -n | tail -1)
+            mostfreq_word=$(echo "$mostfreq" | awk '{ print $2 }')
+            mostfreq_count=$(echo "$mostfreq" | awk '{ print $1 }')
             [[ $words -eq 1 ]] && ess="" || ess="s"
-            inform "OCR complete.  Found ^b${word}^p word${ess}"
+            [[ $mostfreq_count -eq 1 ]] && occ_ess="" || occ_ess="s"
+
+            inform "Found ^b${words}^p word${ess}, most frequent was ^b${mostfreq_word}^p with ^b${mostfreq_count}^p occurence${occ_ess}."
+echo "$allwords" >/tmp/a
         else
-            inform "OCR complete.  Install pdftotext for word counts."
+            warn "pdftotext not installed - skipping word count"
         fi
     else
-        error "OCR failed"
+        fail
     fi
 fi