src/backend/catalog/genbki.sh

   1 #! /bin/sh
   2 #-------------------------------------------------------------------------
   3 #
   4 # genbki.sh--
   5 #    shell script which generates .bki files from specially formatted .h
   6 #    files.  These .bki files are used to initialize the postgres template
   7 #    database.
   8 #
   9 # Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
  10 # Portions Copyright (c) 1994, Regents of the University of California
  11 #
  12 #
  13 # IDENTIFICATION
  14 #    $PostgreSQL$
  15 #
  16 # NOTES
  17 #    non-essential whitespace is removed from the generated file.
  18 #    if this is ever a problem, then the sed script at the very
  19 #    end can be changed into another awk script or something smarter.
  20 #
  21 #-------------------------------------------------------------------------
  22
  23 : ${AWK='awk'}
  24
  25 CMDNAME=`basename $0`
  26
  27 INCLUDE_DIRS=
  28 OUTPUT_PREFIX=
  29 INFILES=
  30 major_version=
  31
  32 #
  33 # Process command line switches.
  34 #
  35 while [ $# -gt 0 ]
  36 do
  37     case $1 in
  38         -I)
  39             INCLUDE_DIRS="$INCLUDE_DIRS $2"
  40             shift;;
  41         -I*)
  42             arg=`echo $1 | sed -e 's/^-I//'`
  43             INCLUDE_DIRS="$INCLUDE_DIRS $arg"
  44             ;;
  45         -o)
  46             OUTPUT_PREFIX="$2"
  47             shift;;
  48         -o*)
  49             OUTPUT_PREFIX=`echo $1 | sed -e 's/^-o//'`
  50             ;;
  51         --set-version=*)
  52             arg=`expr x"$1" : x"--set-version=\(.*\)"`
  53             major_version=`expr x"$arg" : x'\([0-9][0-9]*\.[0-9][0-9]*\)'`
  54             ;;
  55         --help)
  56             echo "$CMDNAME generates system catalog bootstrapping files."
  57             echo
  58             echo "Usage:"
  59             echo "  $CMDNAME [ -I dir ] --set-version=VERSION -o prefix files..."
  60             echo
  61             echo "Options:"
  62             echo "  -I  path to include files"
  63             echo "  -o  prefix of output files"
  64             echo "  --set-version  PostgreSQL version number for initdb cross-check"
  65             echo
  66             echo "The environment variable AWK determines which Awk program"
  67             echo "to use. The default is \`awk'."
  68             echo
  69             echo "Report bugs to <pgsql-bugs@postgresql.org>."
  70             exit 0
  71             ;;
  72         -*)
  73             echo "$CMDNAME: invalid option: $1"
  74             exit 1
  75             ;;
  76         *)
  77             INFILES="$INFILES $1"
  78             ;;
  79     esac
  80     shift
  81 done
  82
  83 if [ x"$INFILES" = x"" ] ; then
  84     echo "$CMDNAME: no input files" 1>&2
  85     exit 1
  86 fi
  87
  88 if [ x"$OUTPUT_PREFIX" = x"" ] ; then
  89     echo "$CMDNAME: no output prefix specified" 1>&2
  90     exit 1
  91 fi
  92
  93 if [ x"$INCLUDE_DIRS" = x"" ] ; then
  94     echo "$CMDNAME: path to include directory unknown" 1>&2
  95     exit 1
  96 fi
  97
  98 if [ x"$major_version" = x"" ] ; then
  99     echo "$CMDNAME: invalid or no version number specified" 1>&2
 100     exit 1
 101 fi
 102
 103
 104 TMPFILE="genbkitmp$$.c"
 105
 106 trap "rm -f $TMPFILE ${OUTPUT_PREFIX}.bki.$$ ${OUTPUT_PREFIX}.description.$$ ${OUTPUT_PREFIX}.shdescription.$$" 0 1 2 3 15
 107
 108
 109 # CAUTION: be wary about what symbols you substitute into the .bki file here!
 110 # It's okay to substitute things that are expected to be really constant
 111 # within a given Postgres release, such as fixed OIDs.  Do not substitute
 112 # anything that could depend on platform or configuration.  (The right place
 113 # to handle those sorts of things is in initdb.c's bootstrap_template1().)
 114
 115 # Get BOOTSTRAP_SUPERUSERID from catalog/pg_authid.h
 116 for dir in $INCLUDE_DIRS; do
 117     if [ -f "$dir/catalog/pg_authid.h" ]; then
 118         BOOTSTRAP_SUPERUSERID=`grep '^#define[  ]*BOOTSTRAP_SUPERUSERID' $dir/catalog/pg_authid.h | $AWK '{ print $3 }'`
 119         break
 120     fi
 121 done
 122
 123 # Get PG_CATALOG_NAMESPACE from catalog/pg_namespace.h
 124 for dir in $INCLUDE_DIRS; do
 125     if [ -f "$dir/catalog/pg_namespace.h" ]; then
 126         PG_CATALOG_NAMESPACE=`grep '^#define[   ]*PG_CATALOG_NAMESPACE' $dir/catalog/pg_namespace.h | $AWK '{ print $3 }'`
 127         break
 128     fi
 129 done
 130
 131 touch ${OUTPUT_PREFIX}.description.$$
 132 touch ${OUTPUT_PREFIX}.shdescription.$$
 133
 134 # ----------------
 135 #       Strip comments and other trash from .h
 136 #
 137 #       Put multi-line start/end comments on a separate line
 138 #
 139 #       Rename datatypes that have different names in .h files than in SQL
 140 #
 141 #       Substitute values of configuration constants
 142 # ----------------
 143 #
 144 cat $INFILES | \
 145 sed -e 's;/\*.*\*/;;g' \
 146     -e 's;/\*;\
 147 /*\
 148 ;g' \
 149     -e 's;\*/;\
 150 */\
 151 ;g' | # we must run a new sed here to see the newlines we added
 152 sed -e "s/;[    ]*$//g" \
 153     -e "s/^[    ]*//" \
 154     -e "s/[     ]Oid/ oid/g" \
 155     -e "s/^Oid/oid/g" \
 156     -e "s/(Oid/(oid/g" \
 157     -e "s/[     ]NameData/ name/g" \
 158     -e "s/^NameData/name/g" \
 159     -e "s/(NameData/(name/g" \
 160     -e "s/[     ]TransactionId/ xid/g" \
 161     -e "s/^TransactionId/xid/g" \
 162     -e "s/(TransactionId/(xid/g" \
 163     -e "s/PGUID/$BOOTSTRAP_SUPERUSERID/g" \
 164     -e "s/PGNSP/$PG_CATALOG_NAMESPACE/g" \
 165 | $AWK '
 166 # ----------------
 167 #       now use awk to process remaining .h file..
 168 #
 169 #       nc is the number of catalogs
 170 #       inside is a variable set to 1 when we are scanning the
 171 #          contents of a catalog definition.
 172 #       reln_open is a flag indicating when we are processing DATA lines.
 173 #          (i.e. have a relation open and need to close it)
 174 #       oid is the most recently seen oid, or 0 if none in the last DATA line.
 175 # ----------------
 176 BEGIN {
 177         inside = 0;
 178         bootstrap = "";
 179         shared_relation = "";
 180         without_oids = "";
 181         nc = 0;
 182         reln_open = 0;
 183         comment_level = 0;
 184         oid = 0;
 185 }
 186
 187 # ----------------
 188 # Anything in a /* .. */ block should be ignored.
 189 # Blank lines also go.
 190 # Note that any /* */ comment on a line by itself was removed from the line
 191 # by the sed above.
 192 # ----------------
 193 /^\/\*/           { comment_level += 1; next; }
 194 /^\*\//           { comment_level -= 1; next; }
 195 comment_level > 0 { next; }
 196
 197 /^[     ]*$/      { next; }
 198
 199 # ----------------
 200 #       DATA() statements are basically passed right through after
 201 #       stripping off the DATA( and the ) on the end.
 202 #       Remember the OID for use by DESCR() and SHDESCR().
 203 # ----------------
 204 /^DATA\(/ {
 205         data = substr($0, 6, length($0) - 6);
 206         oid = 0;
 207         nf = split(data, datafields);
 208         if (nf >= 4 && datafields[1] == "insert" && datafields[2] == "OID" && datafields[3] == "=")
 209         {
 210                 oid = datafields[4];
 211         }
 212         print data;
 213         next;
 214 }
 215
 216 /^DESCR\(/ {
 217         if (oid != 0)
 218         {
 219                 data = substr($0, 8, length($0) - 9);
 220                 if (data != "")
 221                         printf "%d\t%s\t0\t%s\n", oid, catalog, data >>descriptionfile;
 222         }
 223         next;
 224 }
 225
 226 /^SHDESCR\(/ {
 227         if (oid != 0)
 228         {
 229                 data = substr($0, 10, length($0) - 11);
 230                 if (data != "")
 231                         printf "%d\t%s\t%s\n", oid, catalog, data >>shdescriptionfile;
 232         }
 233         next;
 234 }
 235
 236 /^DECLARE_INDEX\(/ {
 237 # ----
 238 #  end any prior catalog data insertions before starting a define index
 239 # ----
 240         if (reln_open == 1) {
 241                 print "close " catalog;
 242                 reln_open = 0;
 243         }
 244
 245         data = substr($0, 15, length($0) - 15);
 246         pos = index(data, ",");
 247         iname = substr(data, 1, pos-1);
 248         data = substr(data, pos+1, length(data)-pos);
 249         pos = index(data, ",");
 250         oid = substr(data, 1, pos-1);
 251         data = substr(data, pos+1, length(data)-pos);
 252
 253         print "declare index " iname " " oid " " data
 254 }
 255
 256 /^DECLARE_UNIQUE_INDEX\(/ {
 257 # ----
 258 #  end any prior catalog data insertions before starting a define unique index
 259 # ----
 260         if (reln_open == 1) {
 261                 print "close " catalog;
 262                 reln_open = 0;
 263         }
 264
 265         data = substr($0, 22, length($0) - 22);
 266         pos = index(data, ",");
 267         iname = substr(data, 1, pos-1);
 268         data = substr(data, pos+1, length(data)-pos);
 269         pos = index(data, ",");
 270         oid = substr(data, 1, pos-1);
 271         data = substr(data, pos+1, length(data)-pos);
 272
 273         print "declare unique index " iname " " oid " " data
 274 }
 275
 276 /^DECLARE_TOAST\(/ {
 277 # ----
 278 #  end any prior catalog data insertions before starting a define toast
 279 # ----
 280         if (reln_open == 1) {
 281                 print "close " catalog;
 282                 reln_open = 0;
 283         }
 284
 285         data = substr($0, 15, length($0) - 15);
 286         pos = index(data, ",");
 287         tname = substr(data, 1, pos-1);
 288         data = substr(data, pos+1, length(data)-pos);
 289         pos = index(data, ",");
 290         toastoid = substr(data, 1, pos-1);
 291         data = substr(data, pos+1, length(data)-pos);
 292         # previous commands already removed the trailing );
 293         indexoid = data;
 294
 295         print "declare toast " toastoid " " indexoid " on " tname
 296 }
 297
 298 /^BUILD_INDICES/        { print "build indices"; }
 299
 300 # ----------------
 301 #       CATALOG() definitions take some more work.
 302 # ----------------
 303 /^CATALOG\(/ {
 304 # ----
 305 #  end any prior catalog data insertions before starting a new one..
 306 # ----
 307         if (reln_open == 1) {
 308                 print "close " catalog;
 309                 reln_open = 0;
 310         }
 311
 312 # ----
 313 #  get the name and properties of the new catalog
 314 # ----
 315         pos = index($1,")");
 316         catalogandoid = substr($1,9,pos-9);
 317         pos = index(catalogandoid, ",");
 318         catalog = substr(catalogandoid, 1, pos-1);
 319         oid = substr(catalogandoid, pos+1, length(catalogandoid)-pos);
 320
 321         if ($0 ~ /BKI_BOOTSTRAP/) {
 322                 bootstrap = "bootstrap ";
 323         }
 324         if ($0 ~ /BKI_SHARED_RELATION/) {
 325                 shared_relation = "shared_relation ";
 326         }
 327         if ($0 ~ /BKI_WITHOUT_OIDS/) {
 328                 without_oids = "without_oids ";
 329         }
 330
 331         i = 1;
 332         inside = 1;
 333         nc++;
 334         next;
 335 }
 336
 337 # ----------------
 338 #       process the columns of the catalog definition
 339 #
 340 #       attname[ x ] contains the attribute name for attribute x
 341 #       atttype[ x ] contains the attribute type fot attribute x
 342 # ----------------
 343 inside == 1 {
 344 # ----
 345 #  ignore a leading brace line..
 346 # ----
 347         if ($1 ~ /\{/)
 348                 next;
 349
 350 # ----
 351 #  if this is the last line, then output the bki catalog stuff.
 352 # ----
 353         if ($1 ~ /}/) {
 354                 print "create " bootstrap shared_relation without_oids catalog " " oid;
 355                 print "\t(";
 356
 357                 for (j=1; j<i-1; j++) {
 358                         print "\t " attname[ j ] " = " atttype[ j ] " ,";
 359                 }
 360                 print "\t " attname[ j ] " = " atttype[ j ] ;
 361                 print "\t)";
 362
 363                 if (bootstrap == "") {
 364                         print "open " catalog;
 365                 }
 366
 367                 i = 1;
 368                 reln_open = 1;
 369                 inside = 0;
 370                 bootstrap = "";
 371                 shared_relation = "";
 372                 without_oids = "";
 373                 next;
 374         }
 375
 376 # ----
 377 #  we are inside the catalog definition, so keep sucking up
 378 #  attribute names and types
 379 # ----
 380         if ($2 ~ /\[.*\]/) {                    # array attribute
 381                 idlen = index($2,"[") - 1;
 382                 atttype[ i ] = $1 "[]";         # variable-length only..
 383                 attname[ i ] = substr($2,1,idlen);
 384         } else {
 385                 atttype[ i ] = $1;
 386                 attname[ i ] = $2;
 387         }
 388         i++;
 389         next;
 390 }
 391
 392 END {
 393         if (reln_open == 1) {
 394                 print "close " catalog;
 395                 reln_open = 0;
 396         }
 397 }
 398 ' "descriptionfile=${OUTPUT_PREFIX}.description.$$" "shdescriptionfile=${OUTPUT_PREFIX}.shdescription.$$" > $TMPFILE || exit
 399
 400 echo "# PostgreSQL $major_version" >${OUTPUT_PREFIX}.bki.$$
 401
 402 sed -e '/^[     ]*$/d' \
 403     -e 's/[     ][      ]*/ /g' $TMPFILE >>${OUTPUT_PREFIX}.bki.$$ || exit
 404
 405 #
 406 # Sanity check: if one of the sed/awk/etc commands fails, we'll probably
 407 # end up with a .bki file that is empty or just a few lines.  Cross-check
 408 # that the files are of reasonable size.  The numbers here are arbitrary,
 409 # but are much smaller than the actual expected sizes as of Postgres 7.2.
 410 #
 411 if [ `wc -c < ${OUTPUT_PREFIX}.bki.$$` -lt 100000 ]; then
 412     echo "$CMDNAME: something seems to be wrong with the .bki file" >&2
 413     exit 1
 414 fi
 415 if [ `wc -c < ${OUTPUT_PREFIX}.description.$$` -lt 10000 ]; then
 416     echo "$CMDNAME: something seems to be wrong with the .description file" >&2
 417     exit 1
 418 fi
 419 if [ `wc -c < ${OUTPUT_PREFIX}.shdescription.$$` -lt 10 ]; then
 420     echo "$CMDNAME: something seems to be wrong with the .shdescription file" >&2
 421     exit 1
 422 fi
 423
 424 # Looks good, commit ...
 425
 426 mv ${OUTPUT_PREFIX}.bki.$$ ${OUTPUT_PREFIX}.bki || exit
 427 mv ${OUTPUT_PREFIX}.description.$$ ${OUTPUT_PREFIX}.description || exit
 428 mv ${OUTPUT_PREFIX}.shdescription.$$ ${OUTPUT_PREFIX}.shdescription || exit
 429
 430 exit 0