Documentation for bootstrap data conversion - Mailing list pgsql-hackers

I felt it was worth spending some extra effort on documentation for
this change, since it's going to impact a lot of future patches.
Accordingly, I've taken John's proposed README text and moved it
into the SGML format, and done a fair amount of editing to extend
the text and bring it all up to date.

The attached patch only touches bki.sgml, but I intend that
src/backend/catalog/README would go away altogether, and we'd
likewise get rid of the separate README.data file that John proposed.
Pretty much all the useful info in those files is here in some form.

One question is whether it's worth renaming bki.sgml to something
more in keeping with its new focus.  I'm inclined not to, because
presumably that would also entail changing the chapter ID to something
else, and I'm not sure how well our documentation website would cope
with that.

John and I are probably both too close to the patch to be able to
review this documentation for clarity and usefulness, so if anyone
else wants to have a look, please comment.

            regards, tom lane

diff --git a/doc/src/sgml/bki.sgml b/doc/src/sgml/bki.sgml
index 33378b4..1e7915e 100644
*** a/doc/src/sgml/bki.sgml
--- b/doc/src/sgml/bki.sgml
***************
*** 1,38 ****
  <!-- doc/src/sgml/bki.sgml -->

  <chapter id="bki">
!  <title><acronym>BKI</acronym> Backend Interface</title>

   <para>
!   Backend Interface (<acronym>BKI</acronym>) files are scripts in a
!   special language that is understood by the
!   <productname>PostgreSQL</productname> backend when running in the
!   <quote>bootstrap</quote> mode.  The bootstrap mode allows system catalogs
!   to be created and filled from scratch, whereas ordinary SQL commands
!   require the catalogs to exist already.
!   <acronym>BKI</acronym> files can therefore be used to create the
!   database system in the first place.  (And they are probably not
!   useful for anything else.)
   </para>

   <para>
!   <application>initdb</application> uses a <acronym>BKI</acronym> file
!   to do part of its job when creating a new database cluster.  The
!   input file used by <application>initdb</application> is created as
!   part of building and installing <productname>PostgreSQL</productname>
!   by a program named <filename>genbki.pl</filename>, which reads some
!   specially formatted C header files in the <filename>src/include/catalog/</filename>
!   directory of the source tree.  The created <acronym>BKI</acronym> file
!   is called <filename>postgres.bki</filename> and is
!   normally installed in the
!   <filename>share</filename> subdirectory of the installation tree.
   </para>

   <para>
!   Related information can be found in the documentation for
!   <application>initdb</application>.
   </para>

   <sect1 id="bki-format">
    <title><acronym>BKI</acronym> File Format</title>

--- 1,641 ----
  <!-- doc/src/sgml/bki.sgml -->

  <chapter id="bki">
!  <title>System Catalog Declarations and Initial Contents</title>

   <para>
!   <productname>PostgreSQL</productname> uses many different system catalogs
!   to keep track of the existence and properties of database objects, such as
!   tables and functions.  Physically there is no difference between a system
!   catalog and a plain user table, but the backend C code knows the structure
!   and properties of each catalog, and can manipulate it directly at a low
!   level.  Thus, for example, it is inadvisable to attempt to alter the
!   structure of a catalog on-the-fly; that would break assumptions built into
!   the C code about how rows of the catalog are laid out.  But developers
!   can change the structure of catalogs in new major versions.
   </para>

   <para>
!   The structures of the catalogs are declared in specially formatted C
!   header files in the <filename>src/include/catalog/</filename> directory of
!   the source tree.  In particular, for each catalog there is a header file
!   named after the catalog (e.g., <filename>pg_class.h</filename>
!   for <structname>pg_class</structname>), which defines the set of columns
!   the catalog has, as well as some other basic properties such as its OID.
!   Other critical files defining the catalog structure
!   include <filename>indexing.h</filename>, which defines the indexes present
!   on all the system catalogs, and <filename>toasting.h</filename>, which
!   defines TOAST tables for catalogs that need one.
   </para>

   <para>
!   Many of the catalogs have initial data that must be loaded into them
!   during the <quote>bootstrap</quote> phase
!   of <application>initdb</application>, to bring the system up to a point
!   where it is capable of executing SQL commands.  (For
!   example, <filename>pg_class.h</filename> must contain an entry for itself,
!   as well as one for each other system catalog and index.)  Much of this
!   initial data is kept in editable form in data files that are also stored
!   in the <filename>src/include/catalog/</filename> directory.  For example,
!   <filename>pg_proc.dat</filename> describes all the initial rows that must
!   be inserted into the <structname>pg_proc</structname> catalog.
   </para>

+  <para>
+   To create the catalog files and load this initial data into them, a
+   backend running in bootstrap mode reads a <acronym>BKI</acronym>
+   (Backend Interface) file containing commands and initial data.
+   The <filename>postgres.bki</filename> file used in this mode is prepared
+   from the aforementioned header and data files, by a Perl script
+   named <filename>genbki.pl</filename>, while building
+   a <productname>PostgreSQL</productname> distribution.
+   Although it's specific to a particular <productname>PostgreSQL</productname>
+   release, <filename>postgres.bki</filename> is platform-independent and is
+   normally installed in the <filename>share</filename> subdirectory of the
+   installation tree.
+  </para>
+
+  <para>
+   <filename>genbki.pl</filename> also produces a derived header file for
+   each catalog, for example <filename>pg_class_d.h</filename> for
+   the <structname>pg_class</structname> catalog.  This file contains
+   automatically-generated macro definitions, and may contain other macros,
+   enum declarations, and so on that can be useful for C code that reads a
+   particular catalog.
+  </para>
+
+  <para>
+   Most developers don't need to be directly concerned with
+   the <acronym>BKI</acronym> file, but almost any nontrivial feature
+   addition in the backend will require modifying the catalog header files
+   and/or initial data files.  The rest of this chapter gives some
+   information about that, and for completeness describes
+   the <acronym>BKI</acronym> file format.
+  </para>
+
+  <sect1 id="system-catalog-declarations">
+   <title>System Catalog Declaration Rules</title>
+
+   <para>
+    The key part of a catalog header file is a C structure definition
+    describing the layout of each row of the catalog.  This begins with
+    a <literal>CATALOG</literal> macro, which so far as the C compiler is
+    concerned is just shorthand for <literal>typedef struct
+    FormData_<replaceable>catalogname</replaceable></literal>.
+    Each field in the struct gives rise to a catalog column.
+    Fields can be annotated using the BKI property macros described
+    in <filename>genbki.h</filename>, to define a default value, mark the
+    field as nullable or not nullable, or specify a lookup rule that allows
+    OID values to be represented symbolically in the
+    corresponding <filename>.dat</filename> file.
+    The <literal>CATALOG</literal> line can also be annotated, with some
+    other BKI property macros described in <filename>genbki.h</filename>, to
+    define other properties of the catalog as a whole, such as whether
+    it has OIDs (by default, it does).
+   </para>
+
+   <para>
+    The system catalog cache code (and most catalog-munging code in general)
+    assumes that the fixed-length portions of all system catalog tuples are
+    in fact present, because it maps this C struct declaration onto them.
+    Thus, all variable-length fields and nullable fields must be placed at
+    the end, and they cannot be accessed as struct fields.
+    For example, if you tried to
+    set <structname>pg_type</structname>.<structfield>typrelid</structfield>
+    to be NULL, it would fail when some piece of code tried to reference
+    <literal>typetup->typrelid</literal> (or worse,
+    <literal>typetup->typelem</literal>, because that follows
+    <structfield>typrelid</structfield>).  This would result in
+    random errors or even segmentation violations.
+   </para>
+
+   <para>
+    As a partial guard against this type of error, variable-length or
+    nullable fields should not be made directly visible to the C compiler.
+    This is accomplished by wrapping them in <literal>#ifdef
+    CATALOG_VARLEN</literal> ... <literal>#endif</literal>.  This prevents C
+    code from carelessly trying to dereference fields that might not be
+    there.  As an independent guard against creating incorrect rows, we
+    require that all columns that should be non-nullable are marked so
+    in <structname>pg_attribute</structname>.  The bootstrap code will
+    automatically mark catalog columns as <literal>NOT NULL</literal>
+    if they are fixed-width and are not preceded by any nullable column.
+    Where this rule is inadequate, you can force correct marking by using
+    <literal>BKI_FORCE_NOT_NULL</literal>
+    and <literal>BKI_FORCE_NULL</literal> annotations as needed.  But note
+    that <literal>NOT NULL</literal> constraints are only enforced in the
+    executor, not against tuples that are generated by random C code,
+    so care is still needed when manually creating or updating catalog rows.
+   </para>
+
+   <para>
+    Frontend code should not include any <structname>pg_xxx.h</structname>
+    catalog header file, as these files may contain C code that won't compile
+    outside the backend.  (Typically, that happens because these files also
+    contain declarations for functions
+    in <filename>src/backend/catalog/</filename> files.)
+    Instead, frontend code may include the corresponding
+    generated <structname>pg_xxx_d.h</structname> header, which will contain
+    OID <literal>#define</literal>s and any other data that might be of use
+    on the client side.  If you want macros or other code in a catalog header
+    to be visible to frontend code, write <literal>#ifdef
+    EXPOSE_TO_CLIENT_CODE</literal> ... <literal>#endif</literal> around that
+    section to instruct <filename>genbki.pl</filename> to copy that section
+    to the <structname>pg_xxx_d.h</structname> header.
+   </para>
+
+   <para>
+    A few of the catalogs are so fundamental that they can't even be created
+    by the <acronym>BKI</acronym> <literal>create</literal> command that's
+    used for most catalogs, because that command needs to write information
+    into these catalogs to describe the new catalog.  These are
+    called <firstterm>bootstrap</firstterm> catalogs, and defining one takes
+    a lot of extra work: you have to manually prepare appropriate entries for
+    them in the pre-loaded contents of <structname>pg_class</structname>
+    and <structname>pg_type</structname>, and those entries will need to be
+    updated for subsequent changes to the catalog's structure.
+    (Bootstrap catalogs also need pre-loaded entries
+    in <structname>pg_attribute</structname>, but
+    fortunately <filename>genbki.pl</filename> handles that chore nowadays.)
+    Avoid making new catalogs be bootstrap catalogs if at all possible.
+   </para>
+  </sect1>
+
+  <sect1 id="system-catalog-initial-data">
+   <title>System Catalog Initial Data</title>
+
+   <para>
+    Each catalog that has any manually-created initial data (some do not)
+    has a corresponding <literal>.dat</literal> file that contains its
+    initial data in an editable format.
+   </para>
+
+   <sect2 id="system-catalog-initial-data-format">
+    <title>Data File Format</title>
+
+    <para>
+     Each <literal>.dat</literal> file contains Perl data structure literals
+     that are simply eval'd to produce an in-memory data structure consisting
+     of an array of hash references, one per catalog row.
+     A slightly modified excerpt from <filename>pg_database.dat</filename>
+     will demonstrate the key features:
+    </para>
+
+ <programlisting>
+ [
+
+ # LC_COLLATE and LC_CTYPE will be replaced at initdb time with user choices
+ # that might contain non-word characters, so we must double-quote them.
+
+ { oid => '1', oid_symbol => 'TemplateDbOid',
+   descr => 'database\'s default template',
+   datname => 'template1', datdba => 'PGUID', encoding => 'ENCODING',
+   datcollate => '"LC_COLLATE"', datctype => '"LC_CTYPE"', datistemplate => 't',
+   datallowconn => 't', datconnlimit => '-1', datlastsysoid => '0',
+   datfrozenxid => '0', datminmxid => '1', dattablespace => '1663',
+   datacl => '_null_' },
+
+ ]
+ </programlisting>
+
+    <para>
+     Points to note:
+    </para>
+
+    <itemizedlist>
+
+     <listitem>
+      <para>
+       The overall file layout is: open square bracket, one or more sets of
+       curly braces each of which represents a catalog row, close square
+       bracket.  Write a comma after each closing curly brace.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       Within each catalog row, write comma-separated
+       <replaceable>key</replaceable> <literal>=></literal>
+       <replaceable>value</replaceable> pairs.  The
+       allowed <replaceable>key</replaceable>s are the names of the catalog's
+       columns, plus the metadata keys <literal>oid</literal>,
+       <literal>oid_symbol</literal>, and <literal>descr</literal>.
+       (The use of <literal>oid</literal> and <literal>oid_symbol</literal>
+       is described in <xref linkend="system-catalog-oid-assignment"/>
+       below.  <literal>descr</literal> supplies a description string for
+       the object, which will be inserted
+       into <structname>pg_description</structname>
+       or <structname>pg_shdescription</structname> as appropriate.)
+       While the metadata keys are optional, the catalog's defined columns
+       must all be provided, except when the catalog's <literal>.h</literal>
+       file specifies a default value for the column.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       All values must be single-quoted.  Escape single quotes used within
+       a value with a backslash.  (Backslashes meant as data need not be
+       doubled, however; this follows Perl's rules for simple quoted
+       literals.)
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       Null values are represented by <literal>_null_</literal>.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       If a value is a macro to be expanded
+       by <application>initdb</application>, it should also contain double
+       quotes as shown above, unless we know that no special characters can
+       appear within the string that will be substituted.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       Comments are preceded by <literal>#</literal>, and must be on their
+       own lines.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       To aid readability, field values that are OIDs of other catalog
+       entries can be represented by names rather than numeric OIDs.
+       This is described in <xref linkend="system-catalog-oid-references"/>
+       below.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       Since hashes are unordered data structures, field order and line
+       layout aren't semantically significant.  However, to maintain a
+       consistent appearance, we set a few rules that are applied by the
+       formatting script <filename>reformat_dat_file.pl</filename>:
+
+       <itemizedlist>
+
+        <listitem>
+         <para>
+          Within each pair of curly braces, the metadata
+          fields <literal>oid</literal>, <literal>oid_symbol</literal>,
+          and <literal>descr</literal> (if present) come first, in that
+          order, then the catalog's own fields appear in their defined order.
+         </para>
+        </listitem>
+
+        <listitem>
+         <para>
+          Newlines are inserted between fields as needed to limit line length
+          to 80 characters, if possible.  A newline is also inserted between
+          the metadata fields and the regular fields.
+         </para>
+        </listitem>
+
+        <listitem>
+         <para>
+          If the catalog's <literal>.h</literal> file specifies a default
+          value for a column, and a data entry has that same
+          value, <filename>reformat_dat_file.pl</filename> will omit it from
+          the data file.  This keeps the data representation compact.
+         </para>
+        </listitem>
+
+        <listitem>
+         <para>
+          <filename>reformat_dat_file.pl</filename> preserves blank lines
+          and comment lines as-is.
+         </para>
+        </listitem>
+
+       </itemizedlist>
+
+       It's recommended to run <filename>reformat_dat_file.pl</filename>
+       before submitting catalog data patches.  For convenience, you can
+       simply change to <filename>src/include/catalog/</filename> and
+       run <literal>make reformat-dat-files</literal>.
+       That script can also be modified to perform bulk editing, as
+       described in <xref linkend="system-catalog-recipes"/> below.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       If you want to add a new method of making the data representation
+       smaller, you must implement it
+       in <filename>reformat_dat_file.pl</filename> and also
+       teach <function>Catalog::ParseData()</function> how to expand the
+       data back into the full representation.
+      </para>
+     </listitem>
+
+    </itemizedlist>
+   </sect2>
+
+   <sect2 id="system-catalog-oid-assignment">
+    <title>OID Assignment</title>
+
+    <para>
+     A catalog row appearing in the initial data can be given a
+     manually-assigned OID by writing an <literal>oid
+     => <replaceable>nnnn</replaceable></literal> metadata field.
+     Furthermore, if an OID is assigned, a C macro for that OID can be
+     created by writing an <literal>oid_symbol
+     => <replaceable>name</replaceable></literal> metadata field.
+    </para>
+
+    <para>
+     Pre-loaded catalog rows must have preassigned OIDs if there are OID
+     references to them in other pre-loaded rows.  A preassigned OID is
+     also needed if the row's OID must be referenced from C code.
+     If neither case applies, the <literal>oid</literal> metadata field can
+     be omitted, in which case the bootstrap code assigns an OID
+     automatically, or leaves it zero in a catalog that has no OIDs.
+     In practice we usually preassign OIDs for all or none of the pre-loaded
+     rows in a given catalog, even if only some of them are actually
+     cross-referenced.
+    </para>
+
+    <para>
+     Writing the actual numeric value of any OID in C code is considered
+     very bad form; always use a macro, instead.  Direct references
+     to <structname>pg_proc</structname> OIDs are common enough that there's
+     a special mechanism to create the necessary macros automatically;
+     see <filename>src/backend/utils/Gen_fmgrtab.pl</filename>.  Similarly
+     — but, for historical reasons, not done the same way —
+     there's an automatic method for creating macros
+     for <structname>pg_type</structname>
+     OIDs.  <literal>oid_symbol</literal> entries are therefore not
+     necessary in those two catalogs.  Likewise, macros for
+     the <structname>pg_class</structname> OIDs of system catalogs and
+     indexes are set up automatically.  For all other system catalogs, you
+     have to manually specify any macros you need
+     via <literal>oid_symbol</literal> entries.
+    </para>
+
+    <para>
+     To find an available OID for a new pre-loaded row, run the
+     script <filename>src/include/catalog/unused_oids</filename>.
+     It prints inclusive ranges of unused OIDs (e.g., the output
+     line <quote>45-900</quote> means OIDs 45 through 900 have not been
+     allocated yet).  Currently, OIDs 1-9999 are reserved for manual
+     assignment; the <filename>unused_oids</filename> script simply looks
+     through the catalog headers and <filename>.dat</filename> files
+     to see which ones do not appear.  You can also use
+     the <filename>duplicate_oids</filename> script to check for mistakes.
+     (That script is run automatically at compile time, and will stop the
+     build if a duplicate is found.)
+    </para>
+
+    <para>
+     The OID counter starts at 10000 at the beginning of a bootstrap run.
+     If a catalog row is in a table that requires OIDs, but no OID was
+     preassigned by an <literal>oid</literal> field, then it will
+     receive an OID of 10000 or above.
+    </para>
+   </sect2>
+
+   <sect2 id="system-catalog-oid-references">
+    <title>OID Reference Lookup</title>
+
+    <para>
+     Cross-references from one initial catalog row to another can be written
+     by just writing the preassigned OID of the referenced row.  But
+     that's error-prone and hard to understand, so for frequently-referenced
+     catalogs, <filename>genbki.pl</filename> provides mechanisms to write
+     symbolic references instead.  Currently this is possible for references
+     to access methods, functions, operators, opclasses, opfamilies, and
+     types.  The rules are as follows:
+    </para>
+
+    <itemizedlist>
+
+     <listitem>
+      <para>
+       Use of symbolic references is enabled in a particular catalog column
+       by attaching <literal>BKI_LOOKUP(<replaceable>lookuprule</replaceable>)</literal>
+       to the column's definition, where <replaceable>lookuprule</replaceable>
+       is <structname>pg_am</structname>, <structname>pg_proc</structname>,
+       <structname>pg_operator</structname>,
+       <structname>pg_opclass</structname>,
+       <structname>pg_opfamily</structname>,
+       or <structname>pg_type</structname>.
+       <literal>BKI_LOOKUP</literal> can be attached to columns of
+       type <type>Oid</type>, <type>regproc</type>, <type>oidvector</type>,
+       or <type>Oid[]</type>; in the latter two cases it implies performing a
+       lookup on each element of the array.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       In such a column, all entries must use the symbolic format except
+       when writing <literal>0</literal> for InvalidOid.  (If the column is
+       declared <type>regproc</type>, you can optionally
+       write <literal>-</literal> instead of <literal>0</literal>.)
+       <filename>genbki.pl</filename> will warn about unrecognized names.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       Access methods are just represented by their names, as are types.
+       Type names must match the referenced <structname>pg_type</structname>
+       entry's <structfield>typname</structfield>; you do not get to use any
+       aliases such as <literal>integer</literal>
+       for <literal>int4</literal>.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       A function can be represented by
+       its <structfield>proname</structfield>, if that is unique among
+       the <filename>pg_proc.dat</filename> entries (this works like regproc
+       input).  Otherwise, write it
+       as <replaceable>proname(argtypename,argtypename,...)</replaceable>,
+       like regprocedure.  The argument type names must be spelled exactly as
+       they are in the <filename>pg_proc.dat</filename> entry's
+       <structfield>proargtypes</structfield> field.  Do not insert any
+       spaces.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       Operators are represented
+       by <replaceable>oprname(lefttype,righttype)</replaceable>,
+       writing the type names exactly as they appear in
+       the <filename>pg_operator.dat</filename>
+       entry's <structfield>oprleft</structfield>
+       and <structfield>oprright</structfield> fields.
+       (Write 0 for the omitted operand of a unary operator.)
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       The names of opclasses and opfamilies are only unique within an
+       access method, so they are represented
+       by <replaceable>access_method_name</replaceable><literal>/</literal><replaceable>object_name</replaceable>.
+      </para>
+     </listitem>
+
+     <listitem>
+      <para>
+       In none of these cases is there any provision for
+       schema-qualification; all objects created during bootstrap are
+       expected to be in the pg_catalog schema.
+      </para>
+     </listitem>
+    </itemizedlist>
+   </sect2>
+
+   <sect2 id="system-catalog-recipes">
+    <title>Recipes for Editing Data Files</title>
+
+    <para>
+     Here are some suggestions about the easiest ways to perform common tasks
+     when updating catalog data files.
+    </para>
+
+    <formalpara>
+     <title>Add a new column with a default to a catalog:</title>
+     <para>
+      Add the column to the header file with
+      a <literal>BKI_DEFAULT(<replaceable>value</replaceable>)</literal>
+      annotation.  The data file need only be adjusted by adding the field
+      in existing rows where a non-default value is needed.
+     </para>
+    </formalpara>
+
+    <formalpara>
+     <title>Add a default value to an existing column that doesn't have
+      one:</title>
+     <para>
+      Add a <literal>BKI_DEFAULT</literal> annotation to the header file,
+      then run <literal>make reformat-dat-files</literal> to remove
+      now-redundant field entries.
+     </para>
+    </formalpara>
+
+    <formalpara>
+     <title>Remove a column, whether it has a default or not:</title>
+     <para>
+      Remove the column from the header, then run <literal>make
+      reformat-dat-files</literal> to remove now-useless field entries.
+     </para>
+    </formalpara>
+
+    <formalpara>
+     <title>Change or remove an existing default value:</title>
+     <para>
+      You cannot simply change the header file, since that will cause the
+      current data to be interpreted incorrectly.  First run <literal>make
+      expand-dat-files</literal> to rewrite the data files with all
+      default values inserted explicitly, then change or remove
+      the <literal>BKI_DEFAULT</literal> annotation, then run <literal>make
+      reformat-dat-files</literal> to remove superfluous fields again.
+     </para>
+    </formalpara>
+
+    <formalpara>
+     <title>Ad-hoc bulk editing:</title>
+     <para>
+      <filename>reformat_dat_file.pl</filename> can be adapted to perform
+      many kinds of bulk changes.  Look for its block comments showing where
+      one-off code can be inserted.  In the following example, we are going
+      to consolidate two boolean fields in <structname>pg_proc</structname>
+      into a char field:
+
+      <orderedlist>
+       <listitem>
+        <para>
+         Add the new column, with a default,
+         to <filename>pg_proc.h</filename>:
+ <programlisting>
+ +    /* see PROKIND_ categories below */
+ +    char        prokind BKI_DEFAULT(f);
+ </programlisting>
+        </para>
+       </listitem>
+
+       <listitem>
+        <para>
+         Create a new script based on <filename>reformat_dat_file.pl</filename>
+         to insert appropriate values on-the-fly:
+ <programlisting>
+ -           # At this point we have the full row in memory as a hash
+ -           # and can do any operations we want. As written, it only
+ -           # removes default values, but this script can be adapted to
+ -           # do one-off bulk-editing.
+ +           # One-off change to migrate to prokind
+ +           # Default has already been filled in by now, so change to other
+ +           # values as appropriate
+ +           if ($values{proisagg} eq 't')
+ +           {
+ +               $values{prokind} = 'a';
+ +           }
+ +           elsif ($values{proiswindow} eq 't')
+ +           {
+ +               $values{prokind} = 'w';
+ +           }
+ </programlisting>
+        </para>
+       </listitem>
+
+       <listitem>
+        <para>
+         Run the new script:
+ <programlisting>
+ $ cd src/include/catalog
+ $ perl -I ../../backend/catalog  rewrite_dat_with_prokind.pl  pg_proc.dat
+ </programlisting>
+         At this point <filename>pg_proc.dat</filename> has all three
+         columns, <structfield>prokind</structfield>,
+         <structfield>proisagg</structfield>,
+         and <structfield>proiswindow</structfield>, though they will appear
+         only in rows where they have non-default values.
+        </para>
+       </listitem>
+
+       <listitem>
+        <para>
+         Remove the old columns from <filename>pg_proc.h</filename>:
+ <programlisting>
+ -    /* is it an aggregate? */
+ -    bool        proisagg BKI_DEFAULT(f);
+ -
+ -    /* is it a window function? */
+ -    bool        proiswindow BKI_DEFAULT(f);
+ </programlisting>
+        </para>
+       </listitem>
+
+       <listitem>
+        <para>
+         Finally, run <literal>make reformat-dat-files</literal> to remove
+         the useless old entries from <filename>pg_proc.dat</filename>.
+        </para>
+       </listitem>
+      </orderedlist>
+
+      For further examples of scripts used for bulk editing, see
+      <filename>convert_oid2name.pl</filename>
+      and <filename>remove_pg_type_oid_symbols.pl</filename> attached to this
+      message:
+      <ulink
url="https://www.postgresql.org/message-id/CAJVSVGX3q=FRRS808Ndk-Sqe96gw3ctSfpz=_dpXN7C+XG256g@mail.gmail.com"></ulink>
+     </para>
+    </formalpara>
+   </sect2>
+  </sect1>
+
   <sect1 id="bki-format">
    <title><acronym>BKI</acronym> File Format</title>

***************
*** 340,346 ****
   </sect1>

   <sect1 id="bki-example">
!   <title>Example</title>

    <para>
     The following sequence of commands will create the
--- 943,949 ----
   </sect1>

   <sect1 id="bki-example">
!   <title>BKI Example</title>

    <para>
     The following sequence of commands will create the