From 602c32207ac0085489dcdd436a6866d35455fc89 Mon Sep 17 00:00:00 2001
From: Dirk Engling <erdgeist@erdgeist.org>
Date: Sat, 23 Feb 2019 23:27:40 +0100
Subject: Add code for database import

---
 src/postprocess/merge_entries.c | 313 ++++++++++++++++++++++++++++++++++++++++
 src/postprocess/postgres.txt    |  41 ++++++
 src/postprocess/postprocess.sh  |  18 +++
 src/postprocess/sort_plz.c      |   4 +-
 4 files changed, 374 insertions(+), 2 deletions(-)
 create mode 100644 src/postprocess/merge_entries.c
 create mode 100644 src/postprocess/postgres.txt
 create mode 100644 src/postprocess/postprocess.sh

diff --git a/src/postprocess/merge_entries.c b/src/postprocess/merge_entries.c
new file mode 100644
index 0000000..3ebfa8c
--- /dev/null
+++ b/src/postprocess/merge_entries.c
@@ -0,0 +1,313 @@
+#include "mystdlib.h"
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <inttypes.h>
+#include <assert.h>
+
+enum { COLUMNS = 15 };
+typedef struct {
+    char *ptr;
+    long rows;
+    long outoff;
+    long flag;
+} entry_t;
+typedef struct {
+    char *ptr;
+    size_t size;
+} outvec_t;
+
+const char *g_year_map[] = {
+"1992_Q2", "1995_Q0", "1996_Q0", "1996_Q1", "1997_Q1", "1997_Q3", "1998_Q1", "1998_Q3", "1999_Q1", "1999_Q3", "2000_Q1", "2000_Q3", "2001_Q1", "2001_Q2", "2001_Q3", "2001_Q4", "2002_Q1",
+"2002_Q3", "2003_Q1", "2003_Q3", "2004_Q1", "2004_Q3", "2005_Q1", "2005_Q3", "2006_Q1", "2006_Q3", "2007_Q1", "2007_Q3", "2008_Q1", "2008_Q3", "2009_Q1", "2009_Q3", "2010_Q1", "2010_Q3",
+"2011_Q1", "2011_Q3", "2012_Q1", "2012_Q3", "2013_Q1", "2013_Q3", "2014_Q1", "2014_Q3", "2015_Q1", "2015_Q3", "2016_Q1", "2016_Q3", "2017_Q1", "2017_Q3", "2018_Q1", "2018_Q3", "2019_Q1",
+0
+};
+
+void SKIP_1_COLUMN(char **ptr) { *ptr = strchr(*ptr, 10) + 1; }
+void SKIP_2_COLUMNS(char **ptr) { SKIP_1_COLUMN(ptr); SKIP_1_COLUMN(ptr); }
+void SKIP_3_COLUMNS(char **ptr) { SKIP_1_COLUMN(ptr); SKIP_1_COLUMN(ptr); SKIP_1_COLUMN(ptr); }
+
+int year_to_offset(const char *year) {
+  const char **y = g_year_map;
+  int off = 0;
+  while (*y) {
+    if (!memcmp(year, *y, 7)) return off;
+    ++off; ++y;
+  }
+  return -1;
+}
+
+
+int
+STRCMP_n (const char *p1, const char *p2)
+{
+  const unsigned char *s1 = (const unsigned char *) p1;
+  const unsigned char *s2 = (const unsigned char *) p2;
+  unsigned char c1, c2;
+  do
+    {
+      c1 = (unsigned char) *s1++;
+      c2 = (unsigned char) *s2++;
+      if (c1 == 10)
+        return c1 - c2;
+    }
+  while (c1 == c2);
+  return c1 - c2;
+}
+
+int compare_entries(entry_t*a, entry_t*b, int *prec) {
+  char *pa = a->ptr, *pb = b->ptr;
+  int col, row, res = 0, nprec = -1;
+
+  /* Multi line entries never match single line entries */
+  if (a->rows != b->rows)
+    return -1;
+
+  /* Assume house number precision first .. unless */
+  if (!memcmp(pa,"2006_Q3",7))
+    *prec = 2;
+  else
+    *prec = 3;
+
+  if (!memcmp(pb,"2006_Q3",7))
+    nprec = 2;
+  else
+    nprec = 3;
+
+  /* Skip year and flags */
+  SKIP_2_COLUMNS(&pa);
+  SKIP_2_COLUMNS(&pb);
+
+  /* Test all columns for identity */
+  for (col=2; col<COLUMNS-1; ++col) {
+    if (!res && STRCMP_n(pa, pb))
+      res = -1;
+    SKIP_1_COLUMN(&pa);
+    SKIP_1_COLUMN(&pb);
+  }
+
+  /* If no coords, downgrade precision */
+  if (*pa == 9) *prec = 1;
+  if (*pb == 9) nprec = 1;
+
+  /* If entries differ, return after precision has been found */
+  if (res) return res;
+
+  /* Only if precision is the same, difference in coordinates
+     is significant.
+  if (*prec == nprec && STRCMP_n(pa, pb))
+    return -1; */
+
+  /* Row 1 has been compared, check the rest of lines */
+  for (row=0; row<a->rows; ++row) {
+
+    /* Skip last row's coordinate columns, year and flags */
+    SKIP_3_COLUMNS(&pa);
+    SKIP_3_COLUMNS(&pb);
+
+    for (col=2; col<COLUMNS-1; ++col) {
+      if (STRCMP_n(pa, pb))
+        return -1;
+      SKIP_1_COLUMN(&pa);
+      SKIP_1_COLUMN(&pb);
+    }
+
+    /* Only if precision is the same, difference in coordinates
+       is significant.
+    if (*prec == nprec && STRCMP_n(pa, pb))
+      return -1; */
+  }
+  return 0;
+}
+
+int sort_me(const void *f_a, const void *f_b) {
+  entry_t *e_a = (entry_t *)f_a; 
+  entry_t *e_b = (entry_t *)f_b; 
+
+  char * pa = (char*)e_a->ptr;
+  char * pb = (char*)e_b->ptr;
+
+  int results[COLUMNS], c;
+
+  if (e_a->rows != e_b->rows)
+    return e_a->rows - e_b->rows;
+
+  for (c = 0; c<COLUMNS; ++c) {
+    results[c] = STRCMP_n(pa, pb);
+    SKIP_1_COLUMN(&pa);
+    SKIP_1_COLUMN(&pb);
+  }
+
+  if (results[10]) return results[10]; /* Vorwahl */
+  if (results[11]) return results[11]; /* Rufnummer */
+  if (results[2])  return results[2];  /* PLZ */
+  if (results[3])  return results[3];  /* Nachname */
+  if (results[4])  return results[4];  /* Vorname */
+  if (results[6])  return results[6];  /* Strasse */
+  if (results[7])  return results[7];  /* Hausnummer */
+  if (results[8])  return results[8];  /* Verweise */
+  if (results[0])  return results[0];  /* Year */
+  return 0;
+}
+
+enum { OUTPUT_BUFFER_SIZE = 1024*1024*128 };
+
+static void do_escape_string(char * s, size_t len) {
+  size_t i;
+
+  putchar('"');
+  for (i=0; i<len; ++i) {
+    switch(s[i]) {
+    case '\\':
+      putchar('\\');
+    case '"':
+      putchar('\\');
+    case ',':
+      putchar('\\');
+    default:
+      break;
+    }
+    putchar(s[i]);
+  }
+  putchar('"');
+}
+
+static void escape_string(char * s, size_t len) {
+  size_t i;
+  if (len == 4 && ( ( s[0] | 0x20 ) == 'n' ) && ( ( s[1] | 0x20 ) == 'u' ) && ( ( s[2] | 0x20 ) == 'l' ) && ( ( s[3] | 0x20 ) == 'l' ) )
+    return do_escape_string(s, len);
+  for (i=0; i<len; ++i)
+    switch(s[i]) {
+      case '{': case '\\': case '}': case '"': case ',': return do_escape_string(s, len);
+    }
+  fwrite(s, len, 1, stdout);
+}
+
+int main(int argc, char **args) {
+  MAP tbuch = map_file(args[1], 1);
+  char *ptr, *start;
+  entry_t * sort_array;
+  outvec_t * out_array;
+  int current = -1, outoff = 0, lines = 1, i, truth = 0, truth_prec = -1;
+  uint64_t year_list = 0, revflag_list = 0, bizflag_list = 0;
+  long flag = 0;
+
+  /* Estimate upper bound for amount of lines */
+  for (i=0; i<tbuch->size; ++i)
+    if (tbuch->addr[i] == 10)
+      ++lines;
+
+  sort_array = (entry_t*)malloc((lines / COLUMNS) * sizeof(entry_t));
+  out_array = (outvec_t*)malloc((lines / COLUMNS) * sizeof(outvec_t));
+
+  ptr = (char*)tbuch->addr;
+  start = ptr;
+
+  while (ptr < (char*)tbuch->addr + tbuch->size) {
+    int c;
+
+    start = ptr;
+
+    /* Look for field terminator */
+    for (c=0; c<COLUMNS; ++c) {
+      char * end = strchr(ptr, 10);
+      if (c==1) {
+        flag = strtoul(ptr, 0, 16);
+        out_array[outoff].ptr = end + 1;
+      }
+      ptr = end + 1;
+    }
+
+    if (flag&2) {
+      assert( current >= 0);
+      sort_array[current].rows++;
+    } else {
+      sort_array[++current].ptr = start;
+      sort_array[current].rows = 0;
+      sort_array[current].outoff = outoff;
+      sort_array[current].flag = flag;
+    }
+    out_array[outoff].size = ptr - out_array[outoff].ptr;
+    outoff++;
+  }
+
+  /* Sort the whole thing */
+  qsort(sort_array, current, sizeof(entry_t), sort_me);
+
+  for (i=0; i<=current; ++i) {
+    int j, dump = 0, prec;
+
+    int year = year_to_offset(sort_array[i].ptr);
+
+    year_list |= 1LL << year;
+    if (sort_array[i].flag & 0x80 ) bizflag_list |= 1LL << year;
+    if (sort_array[i].flag & 0x40 ) revflag_list |= 1LL << year;
+
+    /* The last entry always needs to be dumped, but check if its
+       precision is better than the old truth's
+       The second comparision checks for equality of entries (modulo
+       coordinate mismatch)
+    */
+    if (i == current) {
+      compare_entries(sort_array+i, sort_array+i, &prec);
+      dump = 1;
+    } else if (compare_entries(sort_array+i, sort_array+i+1, &prec))
+      dump = 1;
+
+    /* If this entry's precision is higher than the one of possible
+       earlier matches, then the current entry becomes the truth */
+    if (prec >= truth_prec) {
+      truth = i;
+      truth_prec = prec;
+    }
+
+    if (dump) {
+      printf("%" PRIu64 "\t%" PRIu64 "\t%" PRIu64 "\t", year_list, bizflag_list, revflag_list);
+      for (int c=0; c<COLUMNS-2; ++c) {
+        outvec_t * out = out_array + sort_array[truth].outoff;
+        int started = 0, skipped = 0;
+        for (j=0; j<=sort_array[truth].rows; ++j) {
+          char *s = strchr(out->ptr, 10);
+          size_t len = s - out->ptr;
+          if (!len || out->ptr[0] == 9)
+            skipped++;
+          else {
+            if (!started++)
+              putchar('{');
+            else
+              putchar(',');
+            for (int x=0; x<skipped; ++x) fputs("null,", stdout);
+            if (c != COLUMNS-3)
+              escape_string(out->ptr, len);
+            else {
+              char coords[64], *tab;
+//              memcpy(coords, "POINT(", 6);
+//              memcpy(coords + 6, out->ptr, len);
+//              tab = memchr(coords + 6, 9, len);
+//              if (tab) *tab = ' ';
+//              coords[6+len] = ')';
+//              fwrite(coords, 7 + len, 1, stdout);
+                memcpy(coords, out->ptr, len);
+                tab = memchr(coords, 9, len);
+                if (tab) *tab = ' ';
+                fwrite(coords, len, 1, stdout);
+            }
+            skipped = 0;
+          }
+          out->ptr = s + 1;
+          ++out;
+        }
+        if (started) putchar('}');
+        if (c<COLUMNS-3) putchar(9);
+      }
+      putchar(10);
+
+      truth_prec = -1;
+      year_list = 0;
+      bizflag_list = 0;
+      revflag_list = 0;
+    }
+  }
+}
diff --git a/src/postprocess/postgres.txt b/src/postprocess/postgres.txt
new file mode 100644
index 0000000..d1b4669
--- /dev/null
+++ b/src/postprocess/postgres.txt
@@ -0,0 +1,41 @@
+CREATE TABLE telefonbuch (
+    presence_flag bigint,
+    reverse_flag bigint,
+    biz_flag bigint,
+    zip character varying(64)[] DEFAULT '{}',
+    nachname character varying(256)[] DEFAULT '{}',
+    vorname character varying(192)[] DEFAULT '{}',
+    zusaetze character varying(512)[] DEFAULT '{}',
+    strasse character varying(64)[] DEFAULT '{}',
+    hausnummer character varying(32)[] DEFAULT '{}',
+    verweise text[] DEFAULT '{}',
+    ort character varying(64)[] DEFAULT '{}',
+    vorwahl character varying(10)[] DEFAULT '{}',
+    rufnummer character varying(32)[] DEFAULT '{}',
+    web character varying(256)[] DEFAULT '{}',
+    email character varying(256)[] DEFAULT '{}',
+    coords character varying(32)[] DEFAULT '{}'
+);
+
+COPY Telefonbuch ( presence_flag, reverse_flag, biz_flag, zip, nachname, vorname, zusaetze, strasse, hausnummer, verweise, ort, vorwahl, rufnummer, web, email, coords ) FROM 'telefonbuch.txt' WITH NULL AS 'NULL';
+
+CREATE MATERIALZED VIEW nachname_view AS SELECT id, nachname FROM telefonbuch CROSS JOIN UNNEST(nachname) WITH ORDINALITY u(name, name_id) WITH data;
+
+CREATE INDEX idx_hausnummer ON public.telefonbuch USING btree (hausnummer);
+CREATE INDEX idx_multilinegroup ON public.telefonbuch USING btree (multilinegroup);
+CREATE INDEX idx_nachname ON public.telefonbuch USING btree (nachname);
+CREATE INDEX idx_ort ON public.telefonbuch USING btree (ort);
+CREATE INDEX idx_rufnummer ON public.telefonbuch USING btree (rufnummer);
+CREATE INDEX idx_strasse ON public.telefonbuch USING btree (strasse);
+CREATE INDEX idx_vorname ON public.telefonbuch USING btree (vorname);
+CREATE INDEX idx_vorwahl ON public.telefonbuch USING btree (vorwahl);
+CREATE INDEX idx_zip ON public.telefonbuch USING btree (zip);
+CREATE INDEX idx_zusaetze ON public.telefonbuch USING btree (zusaetze);
+
+CREATE INDEX trgm_idx_nachname ON public.telefonbuch USING gin (nachname gin_trgm_ops);
+CREATE INDEX trgm_idx_rufnummer ON public.telefonbuch USING gin (rufnummer gin_trgm_ops);
+CREATE INDEX trgm_idx_verweise ON public.telefonbuch USING gin (verweise gin_trgm_ops);
+CREATE INDEX trgm_idx_vorname ON public.telefonbuch USING gin (vorname gin_trgm_ops);
+CREATE INDEX trgm_idx_zusaetze ON public.telefonbuch USING gin (zusaetze gin_trgm_ops);
+
+SELECT *, ST_AsText(coords) FROM Telefonbuch WHERE nachname = 'Engling';
diff --git a/src/postprocess/postprocess.sh b/src/postprocess/postprocess.sh
new file mode 100644
index 0000000..cc63a4e
--- /dev/null
+++ b/src/postprocess/postprocess.sh
@@ -0,0 +1,18 @@
+export LANG=C
+export LC_ALL=C
+export LC_CTYPE=C
+
+sudo sysctl kern.maxfilesperproc=65563
+sudo sysctl kern.maxfiles=65563
+ulimit -n 65563
+
+for a in 199[56789]_Q? 20*_Q?; do echo $a; ./sort_plz $a; done
+
+rm -rf sorted
+mkdir sorted
+cd output
+
+for a in *; do echo $a >&2; ../merge_entries $a | iconv -f iso8859-15 -t utf-8 > ../sorted/$a; done
+
+cd ../output
+
diff --git a/src/postprocess/sort_plz.c b/src/postprocess/sort_plz.c
index 26ac9d0..4c30ea3 100644
--- a/src/postprocess/sort_plz.c
+++ b/src/postprocess/sort_plz.c
@@ -17,10 +17,10 @@
 #include <unistd.h>
 #include <err.h>
 
-enum { F_00, F_01, F_10, F_02, F_03, F_04, F_07, F_08, F_09, F_11, F_12, F_13, F_14, F_16, F_COUNT };
+enum { F_00, F_01, F_10, F_02, F_03, F_04, F_07, F_08, F_09, F_11, F_12, F_13, F_14, F_15, F_16, F_COUNT };
 
 static char *g_filenames[] = {
- "00_Jahr", "01_Flags", "10_Postleitzahl", "02_Nachname", "03_Vorname", "04_Zusaetze", "07_Strasse", "08_Hausnummer", "09_Verweise", "11_Ort", "12_Vorwahl", "13_Rufnummer", "14_15_Email_Webadresse", "16_Koordinaten" };
+ "00_Jahr", "01_Flags", "10_Postleitzahl", "02_Nachname", "03_Vorname", "04_Zusaetze", "07_Strasse", "08_Hausnummer", "09_Verweise", "11_Ort", "12_Vorwahl", "13_Rufnummer", "14_Webadresse", "15_Email", "16_Koordinaten" };
 
 typedef struct {
   char plz[8];
-- 
cgit v1.2.3