From 4c3a31b1b03e72e65e080bfcb017ceb9619847a4 Mon Sep 17 00:00:00 2001
From: Dirk Engling <erdgeist@erdgeist.org>
Date: Tue, 2 Jun 2015 19:44:12 +0200
Subject: Cleanup known broken input data, build join.c

---
 makecolumns.sh | 26 ++++++++++++++++++++++++--
 1 file changed, 24 insertions(+), 2 deletions(-)

(limited to 'makecolumns.sh')

diff --git a/makecolumns.sh b/makecolumns.sh
index 8131379..0854b32 100755
--- a/makecolumns.sh
+++ b/makecolumns.sh
@@ -112,6 +112,7 @@ handle_format_version_1() {
     lam 04_Namenszusatz 05_Adresszusatz | tr '\t' ' ' | sed -E s/' +'/' '/g > 04_Zusaetze
     printf "done.\n"
 
+    tidy_streetnames 07_Strasse
 }
 
 handle_format_version_2() {
@@ -158,6 +159,7 @@ handle_format_version_2() {
     lam 04_Namenszusatz 05_Adresszusatz | tr '\t' ' ' | sed -E s/' +'/' '/g > 04_Zusaetze
     printf "done.\n"
 
+    tidy_streetnames 07_Strasse
 }
 
 handle_format_version_3() {
@@ -243,7 +245,7 @@ handle_format_version_3() {
     printf "done.\n"
 
     printf "Normalizing zusaetze ... "
-    lam 04_Namenszusatz 05_Adresszusatz | tr '\t' ' ' | sed -E s/' +'/' '/g > 04_Zusaetze
+    lam 04_Namenszusatz 05_Adresszusatz | tr '\t' ' ' | sed -E -e 's/ +/ /g' -e 's/^ +//g' -e 's/ +$//g' > 04_Zusaetze
     printf "done.\n"
 
     # If street names come in an extra file, extract
@@ -260,10 +262,13 @@ handle_format_version_3() {
         # fix up known broken Strassennamen file
         [ `stat -f %z ${streets}` -eq 1642716 ] && printf '9. Str.\n91. Str.\n91er-Str.\n' >> 99_Strassenname
 
+        tidy_streetnames 99_Strassenname
+
         cut -d ';' -f 1 07_Strassenindex | ${EL} -0x 99_Strassenname > 07_Strasse
         printf "done.\n"
     else
         mv 07_unknown 07_Strasse
+        tidy_streetnames 07_Strasse
     fi
 
     karto=$1/[Dd][Aa][Tt]/[Kk][Aa][Rr][Tt][Oo].[Dd][Aa][Tt]
@@ -306,7 +311,13 @@ handle_format_version_4() {
     find . -name file_\* -delete
     printf "done.\n"
 
-    mv column_0 01_Flags
+    # the 'did not object to inverse search' flag is insane and needs to be reversed
+    if grep -q ^40 column_0; then
+      awk '{ a=substr($0,1,1); printf "%x%x\n",index("5670123cdef89ab4",a)%16 ,substr($0,2,1) }' < column_0 > 01_Flags
+      rm column_0
+    else
+      mv column_0 01_Flags
+    fi
     mv column_1 02_Nachname
     mv column_2 03_Vorname
     mv column_3 04_05_Namenszusatz_Addresszusatz
@@ -318,6 +329,8 @@ handle_format_version_4() {
     mv column_9 13_Rufnummer
     mv column_10 14_15_Email_Webadresse
 
+    tidy_streetnames 99_Strassenname
+
     printf "Looking up street names from indexes ... "
     cut -f 1 07_08_Strassenindex_Hausnummer | ${EL} -0 99_Strassenname > 07_Strasse
     printf "done.\n"
@@ -348,6 +361,15 @@ handle_format_version_4() {
       printf "done.\n"
     fi
     rm file_*
+
+}
+
+tidy_streetnames () {
+  streets=$1
+
+  # Replace any dots at end of line by a single one
+  # finish any str abbreviation without a period with a period
+  sed -Ei '' 's/\.+/./g;s/(S|s)tr( |:)?$/\1tr./' ${streets}
 }
 
 # After function definitions, main() can use them
-- 
cgit v1.2.3