| 1 |
<!-- doc/src/sgml/bki.sgml --> |
| 2 |
|
| 3 |
<chapter id="bki"> |
| 4 |
<title>System Catalog Declarations and Initial Contents</title> |
| 5 |
|
| 6 |
<para> |
| 7 |
<productname>PostgreSQL</productname> uses many different system catalogs |
| 8 |
to keep track of the existence and properties of database objects, such as |
| 9 |
tables and functions. Physically there is no difference between a system |
| 10 |
catalog and a plain user table, but the backend C code knows the structure |
| 11 |
and properties of each catalog, and can manipulate it directly at a low |
| 12 |
level. Thus, for example, it is inadvisable to attempt to alter the |
| 13 |
structure of a catalog on-the-fly; that would break assumptions built into |
| 14 |
the C code about how rows of the catalog are laid out. But the structure |
| 15 |
of the catalogs can change between major versions. |
| 16 |
</para> |
| 17 |
|
| 18 |
<para> |
| 19 |
The structures of the catalogs are declared in specially formatted C |
| 20 |
header files in the <filename>src/include/catalog/</filename> directory of |
| 21 |
the source tree. For each catalog there is a header file |
| 22 |
named after the catalog (e.g., <filename>pg_class.h</filename> |
| 23 |
for <structname>pg_class</structname>), which defines the set of columns |
| 24 |
the catalog has, as well as some other basic properties such as its OID. |
| 25 |
</para> |
| 26 |
|
| 27 |
<para> |
| 28 |
Many of the catalogs have initial data that must be loaded into them |
| 29 |
during the <quote>bootstrap</quote> phase |
| 30 |
of <application>initdb</application>, to bring the system up to a point |
| 31 |
where it is capable of executing SQL commands. (For |
| 32 |
example, <filename>pg_class.h</filename> must contain an entry for itself, |
| 33 |
as well as one for each other system catalog and index.) This |
| 34 |
initial data is kept in editable form in data files that are also stored |
| 35 |
in the <filename>src/include/catalog/</filename> directory. For example, |
| 36 |
<filename>pg_proc.dat</filename> describes all the initial rows that must |
| 37 |
be inserted into the <structname>pg_proc</structname> catalog. |
| 38 |
</para> |
| 39 |
|
| 40 |
<para> |
| 41 |
To create the catalog files and load this initial data into them, a |
| 42 |
backend running in bootstrap mode reads a <acronym>BKI</acronym> |
| 43 |
(Backend Interface) file containing commands and initial data. |
| 44 |
The <filename>postgres.bki</filename> file used in this mode is prepared |
| 45 |
from the aforementioned header and data files, while building |
| 46 |
a <productname>PostgreSQL</productname> distribution, by a Perl script |
| 47 |
named <filename>genbki.pl</filename>. |
| 48 |
Although it's specific to a particular <productname>PostgreSQL</productname> |
| 49 |
release, <filename>postgres.bki</filename> is platform-independent and is |
| 50 |
installed in the <filename>share</filename> subdirectory of the |
| 51 |
installation tree. |
| 52 |
</para> |
| 53 |
|
| 54 |
<para> |
| 55 |
<filename>genbki.pl</filename> also produces a derived header file for |
| 56 |
each catalog, for example <filename>pg_class_d.h</filename> for |
| 57 |
the <structname>pg_class</structname> catalog. This file contains |
| 58 |
automatically-generated macro definitions, and may contain other macros, |
| 59 |
enum declarations, and so on that can be useful for client C code that |
| 60 |
reads a particular catalog. |
| 61 |
</para> |
| 62 |
|
| 63 |
<para> |
| 64 |
Most PostgreSQL developers don't need to be directly concerned with |
| 65 |
the <acronym>BKI</acronym> file, but almost any nontrivial feature |
| 66 |
addition in the backend will require modifying the catalog header files |
| 67 |
and/or initial data files. The rest of this chapter gives some |
| 68 |
information about that, and for completeness describes |
| 69 |
the <acronym>BKI</acronym> file format. |
| 70 |
</para> |
| 71 |
|
| 72 |
<sect1 id="system-catalog-declarations"> |
| 73 |
<title>System Catalog Declaration Rules</title> |
| 74 |
|
| 75 |
<para> |
| 76 |
The key part of a catalog header file is a C structure definition |
| 77 |
describing the layout of each row of the catalog. This begins with |
| 78 |
a <literal>CATALOG</literal> macro, which so far as the C compiler is |
| 79 |
concerned is just shorthand for <literal>typedef struct |
| 80 |
FormData_<replaceable>catalogname</replaceable></literal>. |
| 81 |
Each field in the struct gives rise to a catalog column. |
| 82 |
Fields can be annotated using the BKI property macros described |
| 83 |
in <filename>genbki.h</filename>, for example to define a default value |
| 84 |
for a field or mark it as nullable or not nullable. |
| 85 |
The <literal>CATALOG</literal> line can also be annotated, with some |
| 86 |
other BKI property macros described in <filename>genbki.h</filename>, to |
| 87 |
define other properties of the catalog as a whole, such as whether |
| 88 |
it is a shared relation. |
| 89 |
</para> |
| 90 |
|
| 91 |
<para> |
| 92 |
The system catalog cache code (and most catalog-munging code in general) |
| 93 |
assumes that the fixed-length portions of all system catalog tuples are |
| 94 |
in fact present, because it maps this C struct declaration onto them. |
| 95 |
Thus, all variable-length fields and nullable fields must be placed at |
| 96 |
the end, and they cannot be accessed as struct fields. |
| 97 |
For example, if you tried to |
| 98 |
set <structname>pg_type</structname>.<structfield>typrelid</structfield> |
| 99 |
to be NULL, it would fail when some piece of code tried to reference |
| 100 |
<literal>typetup->typrelid</literal> (or worse, |
| 101 |
<literal>typetup->typelem</literal>, because that follows |
| 102 |
<structfield>typrelid</structfield>). This would result in |
| 103 |
random errors or even segmentation violations. |
| 104 |
</para> |
| 105 |
|
| 106 |
<para> |
| 107 |
As a partial guard against this type of error, variable-length or |
| 108 |
nullable fields should not be made directly visible to the C compiler. |
| 109 |
This is accomplished by wrapping them in <literal>#ifdef |
| 110 |
CATALOG_VARLEN</literal> ... <literal>#endif</literal> (where |
| 111 |
<literal>CATALOG_VARLEN</literal> is a symbol that is never defined). |
| 112 |
This prevents C code from carelessly trying to access fields that might |
| 113 |
not be there or might be at some other offset. |
| 114 |
As an independent guard against creating incorrect rows, we |
| 115 |
require all columns that should be non-nullable to be marked so |
| 116 |
in <structname>pg_attribute</structname>. The bootstrap code will |
| 117 |
automatically mark catalog columns as <literal>NOT NULL</literal> |
| 118 |
if they are fixed-width and are not preceded by any nullable or |
| 119 |
variable-width column. |
| 120 |
Where this rule is inadequate, you can force correct marking by using |
| 121 |
<literal>BKI_FORCE_NOT_NULL</literal> |
| 122 |
and <literal>BKI_FORCE_NULL</literal> annotations as needed. |
| 123 |
</para> |
| 124 |
|
| 125 |
<para> |
| 126 |
Frontend code should not include any <filename>pg_xxx.h</filename> |
| 127 |
catalog header file, as these files may contain C code that won't compile |
| 128 |
outside the backend. (Typically, that happens because these files also |
| 129 |
contain declarations for functions |
| 130 |
in <filename>src/backend/catalog/</filename> files.) |
| 131 |
Instead, frontend code may include the corresponding |
| 132 |
generated <filename>pg_xxx_d.h</filename> header, which will contain |
| 133 |
OID <literal>#define</literal>s and any other data that might be of use |
| 134 |
on the client side. If you want macros or other code in a catalog header |
| 135 |
to be visible to frontend code, write <literal>#ifdef |
| 136 |
EXPOSE_TO_CLIENT_CODE</literal> ... <literal>#endif</literal> around that |
| 137 |
section to instruct <filename>genbki.pl</filename> to copy that section |
| 138 |
to the <filename>pg_xxx_d.h</filename> header. |
| 139 |
</para> |
| 140 |
|
| 141 |
<para> |
| 142 |
A few of the catalogs are so fundamental that they can't even be created |
| 143 |
by the <acronym>BKI</acronym> <literal>create</literal> command that's |
| 144 |
used for most catalogs, because that command needs to write information |
| 145 |
into these catalogs to describe the new catalog. These are |
| 146 |
called <firstterm>bootstrap</firstterm> catalogs, and defining one takes |
| 147 |
a lot of extra work: you have to manually prepare appropriate entries for |
| 148 |
them in the pre-loaded contents of <structname>pg_class</structname> |
| 149 |
and <structname>pg_type</structname>, and those entries will need to be |
| 150 |
updated for subsequent changes to the catalog's structure. |
| 151 |
(Bootstrap catalogs also need pre-loaded entries |
| 152 |
in <structname>pg_attribute</structname>, but |
| 153 |
fortunately <filename>genbki.pl</filename> handles that chore nowadays.) |
| 154 |
Avoid making new catalogs be bootstrap catalogs if at all possible. |
| 155 |
</para> |
| 156 |
</sect1> |
| 157 |
|
| 158 |
<sect1 id="system-catalog-initial-data"> |
| 159 |
<title>System Catalog Initial Data</title> |
| 160 |
|
| 161 |
<para> |
| 162 |
Each catalog that has any manually-created initial data (some do not) |
| 163 |
has a corresponding <literal>.dat</literal> file that contains its |
| 164 |
initial data in an editable format. |
| 165 |
</para> |
| 166 |
|
| 167 |
<sect2 id="system-catalog-initial-data-format"> |
| 168 |
<title>Data File Format</title> |
| 169 |
|
| 170 |
<para> |
| 171 |
Each <literal>.dat</literal> file contains Perl data structure literals |
| 172 |
that are simply eval'd to produce an in-memory data structure consisting |
| 173 |
of an array of hash references, one per catalog row. |
| 174 |
A slightly modified excerpt from <filename>pg_database.dat</filename> |
| 175 |
will demonstrate the key features: |
| 176 |
</para> |
| 177 |
|
| 178 |
<!-- The "slight modification" is the apostrophe in the description. --> |
| 179 |
<programlisting><![CDATA[ |
| 180 |
[ |
| 181 |
|
| 182 |
# A comment could appear here. |
| 183 |
{ oid => '1', oid_symbol => 'Template1DbOid', |
| 184 |
descr => 'database\'s default template', |
| 185 |
datname => 'template1', encoding => 'ENCODING', |
| 186 |
datlocprovider => 'LOCALE_PROVIDER', datistemplate => 't', |
| 187 |
datallowconn => 't', dathasloginevt => 'f', datconnlimit => '-1', datfrozenxid => '0', |
| 188 |
datminmxid => '1', dattablespace => 'pg_default', datcollate => 'LC_COLLATE', |
| 189 |
datctype => 'LC_CTYPE', datlocale => 'DATLOCALE', datacl => '_null_' }, |
| 190 |
|
| 191 |
] |
| 192 |
]]></programlisting> |
| 193 |
|
| 194 |
<para> |
| 195 |
Points to note: |
| 196 |
</para> |
| 197 |
|
| 198 |
<itemizedlist> |
| 199 |
|
| 200 |
<listitem> |
| 201 |
<para> |
| 202 |
The overall file layout is: open square bracket, one or more sets of |
| 203 |
curly braces each of which represents a catalog row, close square |
| 204 |
bracket. Write a comma after each closing curly brace. |
| 205 |
</para> |
| 206 |
</listitem> |
| 207 |
|
| 208 |
<listitem> |
| 209 |
<para> |
| 210 |
Within each catalog row, write comma-separated |
| 211 |
<replaceable>key</replaceable> <literal>=></literal> |
| 212 |
<replaceable>value</replaceable> pairs. The |
| 213 |
allowed <replaceable>key</replaceable>s are the names of the catalog's |
| 214 |
columns, plus the metadata keys <literal>oid</literal>, |
| 215 |
<literal>oid_symbol</literal>, |
| 216 |
<literal>array_type_oid</literal>, and <literal>descr</literal>. |
| 217 |
(The use of <literal>oid</literal> and <literal>oid_symbol</literal> |
| 218 |
is described in <xref linkend="system-catalog-oid-assignment"/> below, |
| 219 |
while <literal>array_type_oid</literal> is described in |
| 220 |
<xref linkend="system-catalog-auto-array-types"/>. |
| 221 |
<literal>descr</literal> supplies a description string for the object, |
| 222 |
which will be inserted into <structname>pg_description</structname> |
| 223 |
or <structname>pg_shdescription</structname> as appropriate.) |
| 224 |
While the metadata keys are optional, the catalog's defined columns |
| 225 |
must all be provided, except when the catalog's <literal>.h</literal> |
| 226 |
file specifies a default value for the column. |
| 227 |
(In the example above, the <structfield>datdba</structfield> field has |
| 228 |
been omitted because <filename>pg_database.h</filename> supplies a |
| 229 |
suitable default value for it.) |
| 230 |
</para> |
| 231 |
</listitem> |
| 232 |
|
| 233 |
<listitem> |
| 234 |
<para> |
| 235 |
All values must be single-quoted. Escape single quotes used within a |
| 236 |
value with a backslash. Backslashes meant as data can, but need not, |
| 237 |
be doubled; this follows Perl's rules for simple quoted literals. |
| 238 |
Note that backslashes appearing as data will be treated as escapes by |
| 239 |
the bootstrap scanner, according to the same rules as for escape string |
| 240 |
constants (see <xref linkend="sql-syntax-strings-escape"/>); for |
| 241 |
example <literal>\t</literal> converts to a tab character. If you |
| 242 |
actually want a backslash in the final value, you will need to write |
| 243 |
four of them: Perl strips two, leaving <literal>\\</literal> for the |
| 244 |
bootstrap scanner to see. |
| 245 |
</para> |
| 246 |
</listitem> |
| 247 |
|
| 248 |
<listitem> |
| 249 |
<para> |
| 250 |
Null values are represented by <literal>_null_</literal>. |
| 251 |
(Note that there is no way to create a value that is just that |
| 252 |
string.) |
| 253 |
</para> |
| 254 |
</listitem> |
| 255 |
|
| 256 |
<listitem> |
| 257 |
<para> |
| 258 |
Comments are preceded by <literal>#</literal>, and must be on their |
| 259 |
own lines. |
| 260 |
</para> |
| 261 |
</listitem> |
| 262 |
|
| 263 |
<listitem> |
| 264 |
<para> |
| 265 |
Field values that are OIDs of other catalog entries should be |
| 266 |
represented by symbolic names rather than actual numeric OIDs. |
| 267 |
(In the example above, <structfield>dattablespace</structfield> |
| 268 |
contains such a reference.) |
| 269 |
This is described in <xref linkend="system-catalog-oid-references"/> |
| 270 |
below. |
| 271 |
</para> |
| 272 |
</listitem> |
| 273 |
|
| 274 |
<listitem> |
| 275 |
<para> |
| 276 |
Since hashes are unordered data structures, field order and line |
| 277 |
layout aren't semantically significant. However, to maintain a |
| 278 |
consistent appearance, we set a few rules that are applied by the |
| 279 |
formatting script <filename>reformat_dat_file.pl</filename>: |
| 280 |
|
| 281 |
<itemizedlist> |
| 282 |
|
| 283 |
<listitem> |
| 284 |
<para> |
| 285 |
Within each pair of curly braces, the metadata |
| 286 |
fields <literal>oid</literal>, <literal>oid_symbol</literal>, |
| 287 |
<literal>array_type_oid</literal>, and <literal>descr</literal> |
| 288 |
(if present) come first, in that order, then the catalog's own |
| 289 |
fields appear in their defined order. |
| 290 |
</para> |
| 291 |
</listitem> |
| 292 |
|
| 293 |
<listitem> |
| 294 |
<para> |
| 295 |
Newlines are inserted between fields as needed to limit line length |
| 296 |
to 80 characters, if possible. A newline is also inserted between |
| 297 |
the metadata fields and the regular fields. |
| 298 |
</para> |
| 299 |
</listitem> |
| 300 |
|
| 301 |
<listitem> |
| 302 |
<para> |
| 303 |
If the catalog's <literal>.h</literal> file specifies a default |
| 304 |
value for a column, and a data entry has that same |
| 305 |
value, <filename>reformat_dat_file.pl</filename> will omit it from |
| 306 |
the data file. This keeps the data representation compact. |
| 307 |
</para> |
| 308 |
</listitem> |
| 309 |
|
| 310 |
<listitem> |
| 311 |
<para> |
| 312 |
<filename>reformat_dat_file.pl</filename> preserves blank lines |
| 313 |
and comment lines as-is. |
| 314 |
</para> |
| 315 |
</listitem> |
| 316 |
|
| 317 |
</itemizedlist> |
| 318 |
|
| 319 |
It's recommended to run <filename>reformat_dat_file.pl</filename> |
| 320 |
before submitting catalog data patches. For convenience, you can |
| 321 |
simply change to <filename>src/include/catalog/</filename> and |
| 322 |
run <literal>make reformat-dat-files</literal>. |
| 323 |
</para> |
| 324 |
</listitem> |
| 325 |
|
| 326 |
<listitem> |
| 327 |
<para> |
| 328 |
If you want to add a new method of making the data representation |
| 329 |
smaller, you must implement it |
| 330 |
in <filename>reformat_dat_file.pl</filename> and also |
| 331 |
teach <function>Catalog::ParseData()</function> how to expand the |
| 332 |
data back into the full representation. |
| 333 |
</para> |
| 334 |
</listitem> |
| 335 |
|
| 336 |
</itemizedlist> |
| 337 |
</sect2> |
| 338 |
|
| 339 |
<sect2 id="system-catalog-oid-assignment"> |
| 340 |
<title>OID Assignment</title> |
| 341 |
|
| 342 |
<para> |
| 343 |
A catalog row appearing in the initial data can be given a |
| 344 |
manually-assigned OID by writing an <literal>oid |
| 345 |
=> <replaceable>nnnn</replaceable></literal> metadata field. |
| 346 |
Furthermore, if an OID is assigned, a C macro for that OID can be |
| 347 |
created by writing an <literal>oid_symbol |
| 348 |
=> <replaceable>name</replaceable></literal> metadata field. |
| 349 |
</para> |
| 350 |
|
| 351 |
<para> |
| 352 |
Pre-loaded catalog rows must have preassigned OIDs if there are OID |
| 353 |
references to them in other pre-loaded rows. A preassigned OID is |
| 354 |
also needed if the row's OID must be referenced from C code. |
| 355 |
If neither case applies, the <literal>oid</literal> metadata field can |
| 356 |
be omitted, in which case the bootstrap code assigns an OID |
| 357 |
automatically. |
| 358 |
In practice we usually preassign OIDs for all or none of the pre-loaded |
| 359 |
rows in a given catalog, even if only some of them are actually |
| 360 |
cross-referenced. |
| 361 |
</para> |
| 362 |
|
| 363 |
<para> |
| 364 |
Writing the actual numeric value of any OID in C code is considered |
| 365 |
very bad form; always use a macro, instead. Direct references |
| 366 |
to <structname>pg_proc</structname> OIDs are common enough that there's |
| 367 |
a special mechanism to create the necessary macros automatically; |
| 368 |
see <filename>src/backend/utils/Gen_fmgrtab.pl</filename>. Similarly |
| 369 |
— but, for historical reasons, not done the same way — |
| 370 |
there's an automatic method for creating macros |
| 371 |
for <structname>pg_type</structname> |
| 372 |
OIDs. <literal>oid_symbol</literal> entries are therefore not |
| 373 |
necessary in those two catalogs. Likewise, macros for |
| 374 |
the <structname>pg_class</structname> OIDs of system catalogs and |
| 375 |
indexes are set up automatically. For all other system catalogs, you |
| 376 |
have to manually specify any macros you need |
| 377 |
via <literal>oid_symbol</literal> entries. |
| 378 |
</para> |
| 379 |
|
| 380 |
<para> |
| 381 |
To find an available OID for a new pre-loaded row, run the |
| 382 |
script <filename>src/include/catalog/unused_oids</filename>. |
| 383 |
It prints inclusive ranges of unused OIDs (e.g., the output |
| 384 |
line <literal>45-900</literal> means OIDs 45 through 900 have not been |
| 385 |
allocated yet). Currently, OIDs 1–9999 are reserved for manual |
| 386 |
assignment; the <filename>unused_oids</filename> script simply looks |
| 387 |
through the catalog headers and <filename>.dat</filename> files |
| 388 |
to see which ones do not appear. You can also use |
| 389 |
the <filename>duplicate_oids</filename> script to check for mistakes. |
| 390 |
(<filename>genbki.pl</filename> will assign OIDs for any rows that |
| 391 |
didn't get one hand-assigned to them, and it will also detect duplicate |
| 392 |
OIDs at compile time.) |
| 393 |
</para> |
| 394 |
|
| 395 |
<para> |
| 396 |
When choosing OIDs for a patch that is not expected to be committed |
| 397 |
immediately, best practice is to use a group of more-or-less |
| 398 |
consecutive OIDs starting with some random choice in the range |
| 399 |
8000—9999. This minimizes the risk of OID collisions with other |
| 400 |
patches being developed concurrently. To keep the 8000—9999 |
| 401 |
range free for development purposes, after a patch has been committed |
| 402 |
to the master git repository its OIDs should be renumbered into |
| 403 |
available space below that range. Typically, this will be done |
| 404 |
near the end of each development cycle, moving all OIDs consumed by |
| 405 |
patches committed in that cycle at the same time. The script |
| 406 |
<filename>renumber_oids.pl</filename> can be used for this purpose. |
| 407 |
If an uncommitted patch is found to have OID conflicts with some |
| 408 |
recently-committed patch, <filename>renumber_oids.pl</filename> may |
| 409 |
also be useful for recovering from that situation. |
| 410 |
</para> |
| 411 |
|
| 412 |
<para> |
| 413 |
Because of this convention of possibly renumbering OIDs assigned by |
| 414 |
patches, the OIDs assigned by a patch should not be considered stable |
| 415 |
until the patch has been included in an official release. We do not |
| 416 |
change manually-assigned object OIDs once released, however, as that |
| 417 |
would create assorted compatibility problems. |
| 418 |
</para> |
| 419 |
|
| 420 |
<para> |
| 421 |
If <filename>genbki.pl</filename> needs to assign an OID to a catalog |
| 422 |
entry that does not have a manually-assigned OID, it will use a value in |
| 423 |
the range 10000—11999. The server's OID counter is set to 10000 |
| 424 |
at the start of a bootstrap run, so that any objects created on-the-fly |
| 425 |
during bootstrap processing also receive OIDs in this range. (The |
| 426 |
usual OID assignment mechanism takes care of preventing any conflicts.) |
| 427 |
</para> |
| 428 |
|
| 429 |
<para> |
| 430 |
Objects with OIDs below <symbol>FirstUnpinnedObjectId</symbol> (12000) |
| 431 |
are considered <quote>pinned</quote>, preventing them from being |
| 432 |
deleted. (There are a small number of exceptions, which are |
| 433 |
hard-wired into <function>IsPinnedObject()</function>.) |
| 434 |
<application>initdb</application> forces the OID counter up |
| 435 |
to <symbol>FirstUnpinnedObjectId</symbol> as soon as it's ready to |
| 436 |
create unpinned objects. Thus objects created during the later phases |
| 437 |
of <application>initdb</application>, such as objects created while |
| 438 |
running the <filename>information_schema.sql</filename> script, will |
| 439 |
not be pinned, while all objects known |
| 440 |
to <filename>genbki.pl</filename> will be. |
| 441 |
</para> |
| 442 |
|
| 443 |
<para> |
| 444 |
OIDs assigned during normal database operation are constrained to be |
| 445 |
16384 or higher. This ensures that the range 10000—16383 is free |
| 446 |
for OIDs assigned automatically by <filename>genbki.pl</filename> or |
| 447 |
during <application>initdb</application>. These |
| 448 |
automatically-assigned OIDs are not considered stable, and may change |
| 449 |
from one installation to another. |
| 450 |
</para> |
| 451 |
</sect2> |
| 452 |
|
| 453 |
<sect2 id="system-catalog-oid-references"> |
| 454 |
<title>OID Reference Lookup</title> |
| 455 |
|
| 456 |
<para> |
| 457 |
In principle, cross-references from one initial catalog row to another |
| 458 |
could be written just by writing the preassigned OID of the referenced |
| 459 |
row in the referencing field. However, that is against project |
| 460 |
policy, because it is error-prone, hard to read, and subject to |
| 461 |
breakage if a newly-assigned OID is renumbered. Therefore |
| 462 |
<filename>genbki.pl</filename> provides mechanisms to write |
| 463 |
symbolic references instead. |
| 464 |
The rules are as follows: |
| 465 |
</para> |
| 466 |
|
| 467 |
<itemizedlist> |
| 468 |
|
| 469 |
<listitem> |
| 470 |
<para> |
| 471 |
Use of symbolic references is enabled in a particular catalog column |
| 472 |
by attaching <literal>BKI_LOOKUP(<replaceable>lookuprule</replaceable>)</literal> |
| 473 |
to the column's definition, where <replaceable>lookuprule</replaceable> |
| 474 |
is the name of the referenced catalog, e.g., <literal>pg_proc</literal>. |
| 475 |
<literal>BKI_LOOKUP</literal> can be attached to columns of |
| 476 |
type <type>Oid</type>, <type>regproc</type>, <type>oidvector</type>, |
| 477 |
or <type>Oid[]</type>; in the latter two cases it implies performing a |
| 478 |
lookup on each element of the array. |
| 479 |
</para> |
| 480 |
</listitem> |
| 481 |
|
| 482 |
<listitem> |
| 483 |
<para> |
| 484 |
It's also permissible to attach <literal>BKI_LOOKUP(encoding)</literal> |
| 485 |
to integer columns to reference character set encodings, which are |
| 486 |
not currently represented as catalog OIDs, but have a set of values |
| 487 |
known to <filename>genbki.pl</filename>. |
| 488 |
</para> |
| 489 |
</listitem> |
| 490 |
|
| 491 |
<listitem> |
| 492 |
<para> |
| 493 |
In some catalog columns, it's allowed for entries to be zero instead |
| 494 |
of a valid reference. If this is allowed, write |
| 495 |
<literal>BKI_LOOKUP_OPT</literal> instead |
| 496 |
of <literal>BKI_LOOKUP</literal>. Then you can |
| 497 |
write <literal>0</literal> for an entry. (If the column is |
| 498 |
declared <type>regproc</type>, you can optionally |
| 499 |
write <literal>-</literal> instead of <literal>0</literal>.) |
| 500 |
Except for this special case, all entries in |
| 501 |
a <literal>BKI_LOOKUP</literal> column must be symbolic references. |
| 502 |
<filename>genbki.pl</filename> will warn about unrecognized names. |
| 503 |
</para> |
| 504 |
</listitem> |
| 505 |
|
| 506 |
<listitem> |
| 507 |
<para> |
| 508 |
Most kinds of catalog objects are simply referenced by their names. |
| 509 |
Note that type names must exactly match the |
| 510 |
referenced <structname>pg_type</structname> |
| 511 |
entry's <structfield>typname</structfield>; you do not get to use |
| 512 |
any aliases such as <literal>integer</literal> |
| 513 |
for <literal>int4</literal>. |
| 514 |
</para> |
| 515 |
</listitem> |
| 516 |
|
| 517 |
<listitem> |
| 518 |
<para> |
| 519 |
A function can be represented by |
| 520 |
its <structfield>proname</structfield>, if that is unique among |
| 521 |
the <filename>pg_proc.dat</filename> entries (this works like regproc |
| 522 |
input). Otherwise, write it |
| 523 |
as <replaceable>proname(argtypename,argtypename,...)</replaceable>, |
| 524 |
like regprocedure. The argument type names must be spelled exactly as |
| 525 |
they are in the <filename>pg_proc.dat</filename> entry's |
| 526 |
<structfield>proargtypes</structfield> field. Do not insert any |
| 527 |
spaces. |
| 528 |
</para> |
| 529 |
</listitem> |
| 530 |
|
| 531 |
<listitem> |
| 532 |
<para> |
| 533 |
Operators are represented |
| 534 |
by <replaceable>oprname(lefttype,righttype)</replaceable>, |
| 535 |
writing the type names exactly as they appear in |
| 536 |
the <filename>pg_operator.dat</filename> |
| 537 |
entry's <structfield>oprleft</structfield> |
| 538 |
and <structfield>oprright</structfield> fields. |
| 539 |
(Write <literal>0</literal> for the omitted operand of a unary |
| 540 |
operator.) |
| 541 |
</para> |
| 542 |
</listitem> |
| 543 |
|
| 544 |
<listitem> |
| 545 |
<para> |
| 546 |
The names of opclasses and opfamilies are only unique within an |
| 547 |
access method, so they are represented |
| 548 |
by <replaceable>access_method_name</replaceable><literal>/</literal><replaceable>object_name</replaceable>. |
| 549 |
</para> |
| 550 |
</listitem> |
| 551 |
|
| 552 |
<listitem> |
| 553 |
<para> |
| 554 |
In none of these cases is there any provision for |
| 555 |
schema-qualification; all objects created during bootstrap are |
| 556 |
expected to be in the <literal>pg_catalog</literal> schema. |
| 557 |
</para> |
| 558 |
</listitem> |
| 559 |
</itemizedlist> |
| 560 |
|
| 561 |
<para> |
| 562 |
<filename>genbki.pl</filename> resolves all symbolic references while it |
| 563 |
runs, and puts simple numeric OIDs into the emitted BKI file. There is |
| 564 |
therefore no need for the bootstrap backend to deal with symbolic |
| 565 |
references. |
| 566 |
</para> |
| 567 |
|
| 568 |
<para> |
| 569 |
It's desirable to mark OID reference columns |
| 570 |
with <literal>BKI_LOOKUP</literal> or <literal>BKI_LOOKUP_OPT</literal> |
| 571 |
even if the catalog has no initial data that requires lookup. This |
| 572 |
allows <filename>genbki.pl</filename> to record the foreign key |
| 573 |
relationships that exist in the system catalogs. That information is |
| 574 |
used in the regression tests to check for incorrect entries. See also |
| 575 |
the macros <literal>DECLARE_FOREIGN_KEY</literal>, |
| 576 |
<literal>DECLARE_FOREIGN_KEY_OPT</literal>, |
| 577 |
<literal>DECLARE_ARRAY_FOREIGN_KEY</literal>, |
| 578 |
and <literal>DECLARE_ARRAY_FOREIGN_KEY_OPT</literal>, which are |
| 579 |
used to declare foreign key relationships that are too complex |
| 580 |
for <literal>BKI_LOOKUP</literal> (typically, multi-column foreign |
| 581 |
keys). |
| 582 |
</para> |
| 583 |
</sect2> |
| 584 |
|
| 585 |
<sect2 id="system-catalog-auto-array-types"> |
| 586 |
<title>Automatic Creation of Array Types</title> |
| 587 |
|
| 588 |
<para> |
| 589 |
Most scalar data types should have a corresponding array type (that is, |
| 590 |
a standard varlena array type whose element type is the scalar type, and |
| 591 |
which is referenced by the <structfield>typarray</structfield> field of |
| 592 |
the scalar type's <structname>pg_type</structname> |
| 593 |
entry). <filename>genbki.pl</filename> is able to generate |
| 594 |
the <structname>pg_type</structname> entry for the array type |
| 595 |
automatically in most cases. |
| 596 |
</para> |
| 597 |
|
| 598 |
<para> |
| 599 |
To use this facility, just write an <literal>array_type_oid |
| 600 |
=> <replaceable>nnnn</replaceable></literal> metadata field in the |
| 601 |
scalar type's <structname>pg_type</structname> entry, specifying the OID |
| 602 |
to use for the array type. You may then omit |
| 603 |
the <structfield>typarray</structfield> field, since it will be filled |
| 604 |
automatically with that OID. |
| 605 |
</para> |
| 606 |
|
| 607 |
<para> |
| 608 |
The generated array type's name is the scalar type's name with an |
| 609 |
underscore prepended. The array entry's other fields are filled from |
| 610 |
<literal>BKI_ARRAY_DEFAULT(<replaceable>value</replaceable>)</literal> |
| 611 |
annotations in <filename>pg_type.h</filename>, or if there isn't one, |
| 612 |
copied from the scalar type. (There's also a special case |
| 613 |
for <structfield>typalign</structfield>.) Then |
| 614 |
the <structfield>typelem</structfield> |
| 615 |
and <structfield>typarray</structfield> fields of the two entries are |
| 616 |
set to cross-reference each other. |
| 617 |
</para> |
| 618 |
</sect2> |
| 619 |
|
| 620 |
<sect2 id="system-catalog-recipes"> |
| 621 |
<title>Recipes for Editing Data Files</title> |
| 622 |
|
| 623 |
<para> |
| 624 |
Here are some suggestions about the easiest ways to perform common tasks |
| 625 |
when updating catalog data files. |
| 626 |
</para> |
| 627 |
|
| 628 |
<formalpara> |
| 629 |
<title>Add a new column with a default to a catalog:</title> |
| 630 |
<para> |
| 631 |
Add the column to the header file with |
| 632 |
a <literal>BKI_DEFAULT(<replaceable>value</replaceable>)</literal> |
| 633 |
annotation. The data file need only be adjusted by adding the field |
| 634 |
in existing rows where a non-default value is needed. |
| 635 |
</para> |
| 636 |
</formalpara> |
| 637 |
|
| 638 |
<formalpara> |
| 639 |
<title>Add a default value to an existing column that doesn't have |
| 640 |
one:</title> |
| 641 |
<para> |
| 642 |
Add a <literal>BKI_DEFAULT</literal> annotation to the header file, |
| 643 |
then run <literal>make reformat-dat-files</literal> to remove |
| 644 |
now-redundant field entries. |
| 645 |
</para> |
| 646 |
</formalpara> |
| 647 |
|
| 648 |
<formalpara> |
| 649 |
<title>Remove a column, whether it has a default or not:</title> |
| 650 |
<para> |
| 651 |
Remove the column from the header, then run <literal>make |
| 652 |
reformat-dat-files</literal> to remove now-useless field entries. |
| 653 |
</para> |
| 654 |
</formalpara> |
| 655 |
|
| 656 |
<formalpara> |
| 657 |
<title>Change or remove an existing default value:</title> |
| 658 |
<para> |
| 659 |
You cannot simply change the header file, since that will cause the |
| 660 |
current data to be interpreted incorrectly. First run <literal>make |
| 661 |
expand-dat-files</literal> to rewrite the data files with all |
| 662 |
default values inserted explicitly, then change or remove |
| 663 |
the <literal>BKI_DEFAULT</literal> annotation, then run <literal>make |
| 664 |
reformat-dat-files</literal> to remove superfluous fields again. |
| 665 |
</para> |
| 666 |
</formalpara> |
| 667 |
|
| 668 |
<formalpara> |
| 669 |
<title>Ad-hoc bulk editing:</title> |
| 670 |
<para> |
| 671 |
<filename>reformat_dat_file.pl</filename> can be adapted to perform |
| 672 |
many kinds of bulk changes. Look for its block comments showing where |
| 673 |
one-off code can be inserted. In the following example, we are going |
| 674 |
to consolidate two Boolean fields in <structname>pg_proc</structname> |
| 675 |
into a char field: |
| 676 |
|
| 677 |
<orderedlist> |
| 678 |
<listitem> |
| 679 |
<para> |
| 680 |
Add the new column, with a default, |
| 681 |
to <filename>pg_proc.h</filename>: |
| 682 |
<programlisting> |
| 683 |
+ /* see PROKIND_ categories below */ |
| 684 |
+ char prokind BKI_DEFAULT(f); |
| 685 |
</programlisting> |
| 686 |
</para> |
| 687 |
</listitem> |
| 688 |
|
| 689 |
<listitem> |
| 690 |
<para> |
| 691 |
Create a new script based on <filename>reformat_dat_file.pl</filename> |
| 692 |
to insert appropriate values on-the-fly: |
| 693 |
<programlisting> |
| 694 |
- # At this point we have the full row in memory as a hash |
| 695 |
- # and can do any operations we want. As written, it only |
| 696 |
- # removes default values, but this script can be adapted to |
| 697 |
- # do one-off bulk-editing. |
| 698 |
+ # One-off change to migrate to prokind |
| 699 |
+ # Default has already been filled in by now, so change to other |
| 700 |
+ # values as appropriate |
| 701 |
+ if ($values{proisagg} eq 't') |
| 702 |
+ { |
| 703 |
+ $values{prokind} = 'a'; |
| 704 |
+ } |
| 705 |
+ elsif ($values{proiswindow} eq 't') |
| 706 |
+ { |
| 707 |
+ $values{prokind} = 'w'; |
| 708 |
+ } |
| 709 |
</programlisting> |
| 710 |
</para> |
| 711 |
</listitem> |
| 712 |
|
| 713 |
<listitem> |
| 714 |
<para> |
| 715 |
Run the new script: |
| 716 |
<programlisting> |
| 717 |
$ cd src/include/catalog |
| 718 |
$ perl rewrite_dat_with_prokind.pl pg_proc.dat |
| 719 |
</programlisting> |
| 720 |
At this point <filename>pg_proc.dat</filename> has all three |
| 721 |
columns, <structfield>prokind</structfield>, |
| 722 |
<structfield>proisagg</structfield>, |
| 723 |
and <structfield>proiswindow</structfield>, though they will appear |
| 724 |
only in rows where they have non-default values. |
| 725 |
</para> |
| 726 |
</listitem> |
| 727 |
|
| 728 |
<listitem> |
| 729 |
<para> |
| 730 |
Remove the old columns from <filename>pg_proc.h</filename>: |
| 731 |
<programlisting> |
| 732 |
- /* is it an aggregate? */ |
| 733 |
- bool proisagg BKI_DEFAULT(f); |
| 734 |
- |
| 735 |
- /* is it a window function? */ |
| 736 |
- bool proiswindow BKI_DEFAULT(f); |
| 737 |
</programlisting> |
| 738 |
</para> |
| 739 |
</listitem> |
| 740 |
|
| 741 |
<listitem> |
| 742 |
<para> |
| 743 |
Finally, run <literal>make reformat-dat-files</literal> to remove |
| 744 |
the useless old entries from <filename>pg_proc.dat</filename>. |
| 745 |
</para> |
| 746 |
</listitem> |
| 747 |
</orderedlist> |
| 748 |
|
| 749 |
For further examples of scripts used for bulk editing, see |
| 750 |
<filename>convert_oid2name.pl</filename> |
| 751 |
and <filename>remove_pg_type_oid_symbols.pl</filename> attached to this |
| 752 |
message: |
| 753 |
<ulink url="https://www.postgresql.org/message-id/CAJVSVGVX8gXnPm+Xa=DxR7kFYprcQ1tNcCT5D0O3ShfnM6jehA@mail.gmail.com"></ulink> |
| 754 |
</para> |
| 755 |
</formalpara> |
| 756 |
</sect2> |
| 757 |
</sect1> |
| 758 |
|
| 759 |
<sect1 id="bki-format"> |
| 760 |
<title><acronym>BKI</acronym> File Format</title> |
| 761 |
|
| 762 |
<para> |
| 763 |
This section describes how the <productname>PostgreSQL</productname> |
| 764 |
backend interprets <acronym>BKI</acronym> files. This description |
| 765 |
will be easier to understand if the <filename>postgres.bki</filename> |
| 766 |
file is at hand as an example. |
| 767 |
</para> |
| 768 |
|
| 769 |
<para> |
| 770 |
<acronym>BKI</acronym> input consists of a sequence of commands. Commands are made up |
| 771 |
of a number of tokens, depending on the syntax of the command. |
| 772 |
Tokens are usually separated by whitespace, but need not be if |
| 773 |
there is no ambiguity. There is no special command separator; the |
| 774 |
next token that syntactically cannot belong to the preceding |
| 775 |
command starts a new one. (Usually you would put a new command on |
| 776 |
a new line, for clarity.) Tokens can be certain key words, special |
| 777 |
characters (parentheses, commas, etc.), identifiers, numbers, or |
| 778 |
single-quoted strings. Everything is case sensitive. |
| 779 |
</para> |
| 780 |
|
| 781 |
<para> |
| 782 |
Lines starting with <literal>#</literal> are ignored. |
| 783 |
</para> |
| 784 |
|
| 785 |
</sect1> |
| 786 |
|
| 787 |
<sect1 id="bki-commands"> |
| 788 |
<title><acronym>BKI</acronym> Commands</title> |
| 789 |
|
| 790 |
<variablelist> |
| 791 |
<varlistentry> |
| 792 |
<term> |
| 793 |
<literal>create</literal> |
| 794 |
<replaceable class="parameter">tablename</replaceable> |
| 795 |
<replaceable class="parameter">tableoid</replaceable> |
| 796 |
<optional><literal>bootstrap</literal></optional> |
| 797 |
<optional><literal>shared_relation</literal></optional> |
| 798 |
<optional><literal>rowtype_oid</literal> <replaceable>oid</replaceable></optional> |
| 799 |
(<replaceable class="parameter">name1</replaceable> = |
| 800 |
<replaceable class="parameter">type1</replaceable> |
| 801 |
<optional><literal>FORCE NOT NULL</literal> | <literal>FORCE NULL</literal> </optional> <optional>, |
| 802 |
<replaceable class="parameter">name2</replaceable> = |
| 803 |
<replaceable class="parameter">type2</replaceable> |
| 804 |
<optional><literal>FORCE NOT NULL</literal> | <literal>FORCE NULL</literal> </optional>, |
| 805 |
...</optional>) |
| 806 |
</term> |
| 807 |
|
| 808 |
<listitem> |
| 809 |
<para> |
| 810 |
Create a table named <replaceable |
| 811 |
class="parameter">tablename</replaceable>, and having the OID |
| 812 |
<replaceable class="parameter">tableoid</replaceable>, |
| 813 |
with the columns given in parentheses. |
| 814 |
</para> |
| 815 |
|
| 816 |
<para> |
| 817 |
The following column types are supported directly by |
| 818 |
<filename>bootstrap.c</filename>: <type>bool</type>, |
| 819 |
<type>bytea</type>, <type>char</type> (1 byte), |
| 820 |
<type>name</type>, <type>int2</type>, |
| 821 |
<type>int4</type>, <type>regproc</type>, <type>regclass</type>, |
| 822 |
<type>regtype</type>, <type>text</type>, |
| 823 |
<type>oid</type>, <type>tid</type>, <type>xid</type>, |
| 824 |
<type>cid</type>, <type>int2vector</type>, <type>oidvector</type>, |
| 825 |
<type>_int4</type> (array), <type>_text</type> (array), |
| 826 |
<type>_oid</type> (array), <type>_char</type> (array), |
| 827 |
<type>_aclitem</type> (array). Although it is possible to create |
| 828 |
tables containing columns of other types, this cannot be done until |
| 829 |
after <structname>pg_type</structname> has been created and filled with |
| 830 |
appropriate entries. (That effectively means that only these |
| 831 |
column types can be used in bootstrap catalogs, but non-bootstrap |
| 832 |
catalogs can contain any built-in type.) |
| 833 |
</para> |
| 834 |
|
| 835 |
<para> |
| 836 |
When <literal>bootstrap</literal> is specified, |
| 837 |
the table will only be created on disk; nothing is entered into |
| 838 |
<structname>pg_class</structname>, |
| 839 |
<structname>pg_attribute</structname>, etc., for it. Thus the |
| 840 |
table will not be accessible by ordinary SQL operations until |
| 841 |
such entries are made the hard way (with <literal>insert</literal> |
| 842 |
commands). This option is used for creating |
| 843 |
<structname>pg_class</structname> etc. themselves. |
| 844 |
</para> |
| 845 |
|
| 846 |
<para> |
| 847 |
The table is created as shared if <literal>shared_relation</literal> is |
| 848 |
specified. |
| 849 |
The table's row type OID (<structname>pg_type</structname> OID) can optionally |
| 850 |
be specified via the <literal>rowtype_oid</literal> clause; if not specified, |
| 851 |
an OID is automatically generated for it. (The <literal>rowtype_oid</literal> |
| 852 |
clause is useless if <literal>bootstrap</literal> is specified, but it can be |
| 853 |
provided anyway for documentation.) |
| 854 |
</para> |
| 855 |
</listitem> |
| 856 |
</varlistentry> |
| 857 |
|
| 858 |
<varlistentry> |
| 859 |
<term> |
| 860 |
<literal>open</literal> <replaceable class="parameter">tablename</replaceable> |
| 861 |
</term> |
| 862 |
|
| 863 |
<listitem> |
| 864 |
<para> |
| 865 |
Open the table named |
| 866 |
<replaceable class="parameter">tablename</replaceable> |
| 867 |
for insertion of data. Any currently open table is closed. |
| 868 |
</para> |
| 869 |
</listitem> |
| 870 |
</varlistentry> |
| 871 |
|
| 872 |
<varlistentry> |
| 873 |
<term> |
| 874 |
<literal>close</literal> <replaceable class="parameter">tablename</replaceable> |
| 875 |
</term> |
| 876 |
|
| 877 |
<listitem> |
| 878 |
<para> |
| 879 |
Close the open table. The name of the table must be given as a |
| 880 |
cross-check. |
| 881 |
</para> |
| 882 |
</listitem> |
| 883 |
</varlistentry> |
| 884 |
|
| 885 |
<varlistentry> |
| 886 |
<term> |
| 887 |
<literal>insert</literal> <literal>(</literal> <optional><replaceable class="parameter">oid_value</replaceable></optional> <replaceable class="parameter">value1</replaceable> <replaceable class="parameter">value2</replaceable> ... <literal>)</literal> |
| 888 |
</term> |
| 889 |
|
| 890 |
<listitem> |
| 891 |
<para> |
| 892 |
Insert a new row into the open table using <replaceable |
| 893 |
class="parameter">value1</replaceable>, <replaceable |
| 894 |
class="parameter">value2</replaceable>, etc., for its column |
| 895 |
values. |
| 896 |
</para> |
| 897 |
|
| 898 |
<para> |
| 899 |
NULL values can be specified using the special key word |
| 900 |
<literal>_null_</literal>. Values that do not look like |
| 901 |
identifiers or digit strings must be single-quoted. |
| 902 |
(To include a single quote in a value, write it twice. |
| 903 |
Escape-string-style backslash escapes are allowed in the string, too.) |
| 904 |
</para> |
| 905 |
</listitem> |
| 906 |
</varlistentry> |
| 907 |
|
| 908 |
<varlistentry> |
| 909 |
<term> |
| 910 |
<literal>declare</literal> <optional><literal>unique</literal></optional> |
| 911 |
<literal>index</literal> <replaceable class="parameter">indexname</replaceable> |
| 912 |
<replaceable class="parameter">indexoid</replaceable> |
| 913 |
<literal>on</literal> <replaceable class="parameter">tablename</replaceable> |
| 914 |
<literal>using</literal> <replaceable class="parameter">amname</replaceable> |
| 915 |
<literal>(</literal> <replaceable class="parameter">opclass1</replaceable> |
| 916 |
<replaceable class="parameter">name1</replaceable> |
| 917 |
<optional>, ...</optional> <literal>)</literal> |
| 918 |
</term> |
| 919 |
|
| 920 |
<listitem> |
| 921 |
<para> |
| 922 |
Create an index named <replaceable |
| 923 |
class="parameter">indexname</replaceable>, having OID |
| 924 |
<replaceable class="parameter">indexoid</replaceable>, |
| 925 |
on the table named |
| 926 |
<replaceable class="parameter">tablename</replaceable>, using the |
| 927 |
<replaceable class="parameter">amname</replaceable> access |
| 928 |
method. The fields to index are called <replaceable |
| 929 |
class="parameter">name1</replaceable>, <replaceable |
| 930 |
class="parameter">name2</replaceable> etc., and the operator |
| 931 |
classes to use are <replaceable |
| 932 |
class="parameter">opclass1</replaceable>, <replaceable |
| 933 |
class="parameter">opclass2</replaceable> etc., respectively. |
| 934 |
The index file is created and appropriate catalog entries are |
| 935 |
made for it, but the index contents are not initialized by this command. |
| 936 |
</para> |
| 937 |
</listitem> |
| 938 |
</varlistentry> |
| 939 |
|
| 940 |
<varlistentry> |
| 941 |
<term> |
| 942 |
<literal>declare toast</literal> |
| 943 |
<replaceable class="parameter">toasttableoid</replaceable> |
| 944 |
<replaceable class="parameter">toastindexoid</replaceable> |
| 945 |
<literal>on</literal> <replaceable class="parameter">tablename</replaceable> |
| 946 |
</term> |
| 947 |
|
| 948 |
<listitem> |
| 949 |
<para> |
| 950 |
Create a TOAST table for the table named |
| 951 |
<replaceable class="parameter">tablename</replaceable>. |
| 952 |
The TOAST table is assigned OID |
| 953 |
<replaceable class="parameter">toasttableoid</replaceable> |
| 954 |
and its index is assigned OID |
| 955 |
<replaceable class="parameter">toastindexoid</replaceable>. |
| 956 |
As with <literal>declare index</literal>, filling of the index |
| 957 |
is postponed. |
| 958 |
</para> |
| 959 |
</listitem> |
| 960 |
</varlistentry> |
| 961 |
|
| 962 |
<varlistentry> |
| 963 |
<term><literal>build indices</literal></term> |
| 964 |
|
| 965 |
<listitem> |
| 966 |
<para> |
| 967 |
Fill in the indices that have previously been declared. |
| 968 |
</para> |
| 969 |
</listitem> |
| 970 |
</varlistentry> |
| 971 |
</variablelist> |
| 972 |
|
| 973 |
</sect1> |
| 974 |
|
| 975 |
<sect1 id="bki-structure"> |
| 976 |
<title>Structure of the Bootstrap <acronym>BKI</acronym> File</title> |
| 977 |
|
| 978 |
<para> |
| 979 |
The <literal>open</literal> command cannot be used until the tables it uses |
| 980 |
exist and have entries for the table that is to be opened. |
| 981 |
(These minimum tables are <structname>pg_class</structname>, |
| 982 |
<structname>pg_attribute</structname>, <structname>pg_proc</structname>, and |
| 983 |
<structname>pg_type</structname>.) To allow those tables themselves to be filled, |
| 984 |
<literal>create</literal> with the <literal>bootstrap</literal> option implicitly opens |
| 985 |
the created table for data insertion. |
| 986 |
</para> |
| 987 |
|
| 988 |
<para> |
| 989 |
Also, the <literal>declare index</literal> and <literal>declare toast</literal> |
| 990 |
commands cannot be used until the system catalogs they need have been |
| 991 |
created and filled in. |
| 992 |
</para> |
| 993 |
|
| 994 |
<para> |
| 995 |
Thus, the structure of the <filename>postgres.bki</filename> file has to |
| 996 |
be: |
| 997 |
<orderedlist> |
| 998 |
<listitem> |
| 999 |
<para> |
| 1000 |
<literal>create bootstrap</literal> one of the critical tables |
| 1001 |
</para> |
| 1002 |
</listitem> |
| 1003 |
<listitem> |
| 1004 |
<para> |
| 1005 |
<literal>insert</literal> data describing at least the critical tables |
| 1006 |
</para> |
| 1007 |
</listitem> |
| 1008 |
<listitem> |
| 1009 |
<para> |
| 1010 |
<literal>close</literal> |
| 1011 |
</para> |
| 1012 |
</listitem> |
| 1013 |
<listitem> |
| 1014 |
<para> |
| 1015 |
Repeat for the other critical tables. |
| 1016 |
</para> |
| 1017 |
</listitem> |
| 1018 |
<listitem> |
| 1019 |
<para> |
| 1020 |
<literal>create</literal> (without <literal>bootstrap</literal>) a noncritical table |
| 1021 |
</para> |
| 1022 |
</listitem> |
| 1023 |
<listitem> |
| 1024 |
<para> |
| 1025 |
<literal>open</literal> |
| 1026 |
</para> |
| 1027 |
</listitem> |
| 1028 |
<listitem> |
| 1029 |
<para> |
| 1030 |
<literal>insert</literal> desired data |
| 1031 |
</para> |
| 1032 |
</listitem> |
| 1033 |
<listitem> |
| 1034 |
<para> |
| 1035 |
<literal>close</literal> |
| 1036 |
</para> |
| 1037 |
</listitem> |
| 1038 |
<listitem> |
| 1039 |
<para> |
| 1040 |
Repeat for the other noncritical tables. |
| 1041 |
</para> |
| 1042 |
</listitem> |
| 1043 |
<listitem> |
| 1044 |
<para> |
| 1045 |
Define indexes and toast tables. |
| 1046 |
</para> |
| 1047 |
</listitem> |
| 1048 |
<listitem> |
| 1049 |
<para> |
| 1050 |
<literal>build indices</literal> |
| 1051 |
</para> |
| 1052 |
</listitem> |
| 1053 |
</orderedlist> |
| 1054 |
</para> |
| 1055 |
|
| 1056 |
<para> |
| 1057 |
There are doubtless other, undocumented ordering dependencies. |
| 1058 |
</para> |
| 1059 |
</sect1> |
| 1060 |
|
| 1061 |
<sect1 id="bki-example"> |
| 1062 |
<title>BKI Example</title> |
| 1063 |
|
| 1064 |
<para> |
| 1065 |
The following sequence of commands will create the table |
| 1066 |
<literal>test_table</literal> with OID 420, having three columns |
| 1067 |
<literal>oid</literal>, <literal>cola</literal> and <literal>colb</literal> |
| 1068 |
of type <type>oid</type>, <type>int4</type> and <type>text</type>, |
| 1069 |
respectively, and insert two rows into the table: |
| 1070 |
<programlisting> |
| 1071 |
create test_table 420 (oid = oid, cola = int4, colb = text) |
| 1072 |
open test_table |
| 1073 |
insert ( 421 1 'value 1' ) |
| 1074 |
insert ( 422 2 _null_ ) |
| 1075 |
close test_table |
| 1076 |
</programlisting> |
| 1077 |
</para> |
| 1078 |
</sect1> |
| 1079 |
</chapter> |