SQL-Routine
----------------------------------------------------------------------
2005-09-08 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.70.0.
* New code file versions are: Routine.pm v0.70.0 and en.pm v0.38.0.
* Updated external dependencies on Locale::KeyedText to v1.6.0.
* From now on, if the minimum Perl version that this distribution depends on
(currently 5.008001) comes bundled with a satisfactory version of a module
that we have a dependency on, then we will no longer specify any version
number when referring to said module. An example of such is Test::More.
* Updated the DESCRIPTION paragraph of KeyedText.pm that refers to how easy
to install the module is; it now says the module has few external
dependencies, rather than specifying one.
* Converted every distribution file to use indentations composed of space
characters (4 spaces per indentation level) rather than tab characters.
* Updated the 2 Routine.pm private methods _serialize_as_[perl|xml]() to
indent their output using spaces (4 per level) rather than tabs.
* Updated all files containing POD (*.pm, *.pod) to re-order some POD
sections. Now all of the POD is after all of the code, rather than that
being mostly true but for a bit of POD before most of the code. The 'NAME'
POD has been moved to just above the 'SYNOPSIS'. The 'DEPENDENCIES' has been
moved to just above the 'SEE ALSO'. Any 'BUGS' and 'CAVEATS' have been
moved to just below the 'SEE ALSO'. The 'COPYRIGHT AND LICENSE' has been
moved to the very end of the file, except that any 'CREDITS' have been moved
below that. Also updated the ReadMe file to move the 'CREDITS' below the
'COPYRIGHT AND LICENSE'. There were zero changes to the content of any POD
sections, and no new POD sections were added or existing ones removed,
except for those explicitly listed below.
* Updated all files containing POD: Deleted the 'SYNTAX' section, whose
value was dubious, from any files that had one. Renamed 'COPYRIGHT AND
LICENSE' to 'LICENCE AND COPYRIGHT', and added a new 'AUTHOR' section just
above it. Renamed 'CREDITS' to 'ACKNOWLEDGEMENTS' in any files that had
one. COPYRIGHT and CREDITS were likewise renamed in the ReadMe file.
* Updated the POD in all *.pm files: Added a new 'VERSION' section, which
appears just below the 'NAME'. Added a new 'INCOMPATIBILITIES' section,
which appears just below 'DEPENDENCIES'. Renamed and/or merged any 'BUGS'
and 'CAVEATS' sections into a single 'BUGS AND LIMITATIONS' section, and
added that section to any files lacking its predecessors. Rewrote the
'DEPENDENCIES' into a new paragraph-resembling format that specifies where
each dependency is, such as bundled with perl, or bundled with the current
distribution, or available on CPAN; also rewrote the ReadMe file version.
* Updated all *.pm files to add a new external dependency on the 'version'
module, which has core-like functionality but isn't yet bundled with Perl;
all 'our $VERSION' declarations were changed to qv-wrapped three-part format
from floating point format; likewise, updated the *_00_Compile.t file to use
'version', and also changed the tests for expected module versions to be
performed by is() rather than cmp_ok().
* Updated Routine.pm to add a new external dependency on the 'only'
module, which has core-like functionality but isn't yet bundled with Perl;
all "use Foo N.NN" statements were changed to three-part format and now look
like "use only 'Foo' => 'N.N.N-'".
2005-09-01 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.69.
* New code file versions are: Routine.pm v0.69 and en.pm v0.37.
* Removed a long-standing erroneous "=head1" in the ReadMe file, which
caused the search.cpan.org documentation maker to mis-format that file.
* Corresponding changes in Routine.pm, Details.pod, en.pm: Merged the 18
Node interface methods named
[get|clear|set]_[literal|enumerated|node_ref]_attribute[|s]() into the
pre-existing 6 Node interface methods named [get|clear_set]_attribute[|s]();
the newly size-reduced API of SQL::Routine should hopefully be less daunting
to programmers than the old API, even though it is now a bit less
self-documenting. As part of this change, the 6 en.pm user text messages
'SRT_N_METH_ARG_NO_[LIT|ENUM|NREF]_AT_NM' and
'SRT_N_SET_[LIT|ENUM|NREF]_AT_NO_ARG_VAL' were merged into the pre-existing
2 messages 'SRT_N_METH_ARG_NO_AT_NM' and 'SRT_N_SET_AT_NO_ARG_VAL'. There
were also 11 other en.pm messages altered, that were used by private methods
which the old 24 public methods had invoked. There were also 4 en.pm
messages moved upward in the file to just below the PP messages. In
Routine.pm, 9 NodeStorage private methods were slightly altered to use the
aforementioned altered en.pm messages,
_clear_[literal|enumerated|node_ref]_attribute[|s]() and
_set_[literal|enumerated|node_ref]_attribute(). Besides the above, there
were a few trivial related documentation updates in Routine.pm, Details.pod.
* Updated the 2 test suite files t_SRT_Verbose.pm and SRT_10_Identity.t to
replace invocations of the removed methods with invocations of their shorter
counterparts; SRT_10_Identity.t also had a net 2 test blocks removed.
2005-09-01 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.68.
* New code file versions are: Routine.pm v0.68 and en.pm v0.36.
* Added a new main title to this Changes file that matches the ReadMe file.
* Rewrote the documentation section KEEPING UP TO DATE in the ReadMe file.
* Corresponding changes in Routine.pm, Details.pod, en.pm: Deleted the
boolean property "is read only" from the ContainerStorage class, along with
its accessor method is_read_only(); this will be replaced with alternate
functionality called "access controls". Also updated the Routine.pm code
for these 21 methods to remove references to the property:
auto_assert_deferrable_constraints(), auto_set_node_ids(),
may_match_surrogate_node_ids(), Container.delete_node_tree(),
ContainerStorage._new(), move_before_sibling(), NodeStorage._new(),
_delete_node[|_tree](), _set_node_id(),
_[clear|set]_primary_parent_attribute(),
_clear_[literal|enumerated|node_ref]_attribute[|s](),
_set_[literal|enumerated|node_ref]_attribute(). Also removed the 2 en.pm
user text messages 'SRT_[C|N]_METH_ASS_READ_ONLY'.
* Deleted test file t/SRT_10_ReadOnly.t, which is no longer applicable.
* Corresponding changes in Routine.pm, Details.pod, en.pm: Moved these 3
boolean properties from the ContainerStorage class to the Container
interface class: "auto assert deferrable constraints", "auto set node ids",
"may match surrogate node ids"; from now on, setting any of these 3
properties to true will only affect program actions done through their own
Container interface (or child Node interfaces), and not those done through
other interfaces. Updated the Routine.pm code and Details.pod documentation
of the corresponding 3 property accessor methods. Updated the Routine.pm
code for these 6 methods that reference those properties: Container.new(),
new_interface(), _build_node_is_child_or_not(), ContainerStorage._new(),
Node.new(), _set_node_ref_attribute(). Also updated the 2 en.pm user text
messages 'SRT_N_NEW_NODE_NO_ARG_ID' and 'SRT_N_SET_NREF_AT_NO_ALLOW_SID'.
* Large update to Details.pod: Added new main documentation section FAULT
TOLERANCE, DATA INTEGRITY, AND CONCURRENCY; it has 4 sub-sections named
"Structural Matters", "Function or Method Types and Exceptions", "Input
Validation", and "Multiple Thread Concurrency". Removed main documentation
section FAULT TOLERANCE AND MULTI-THREADING SUPPORT; all of its contents
have been moved into the aforementioned new main section, comprising a
majority of all the former's sub-sections except for "Input Validation".
Removed all but the first SYNTAX paragraph; all of that content has been
moved into the aforementioned new main section, mostly becoming "Input
Validation". FAULT ... has a small amount of new material also.
* Updated Routine.pm to fix a bug in _delete_node_tree() and
_delete_node_tree__add_to_candidates(); Perl 5 hashes always stringify their
keys, so we can't use the keys of %candidates as objects; updated
%candidates to store the actual NodeStorage ref in the hash value, and
updated use code to iterate values instead of keys.
* Corresponding minor bug fix in the SYNOPSIS of Routine.pm and the code of
t_SRT_Synopsis.pm: corrected 3 syntax errors where members of 3 key/value
pairs were separated from their partners by ',' rather than '=>'.
* Added new test file t/SRT_10_Identity.t, to test the new_interface() and
get_self_id() functions that were added in release 0.66, plus other aspects
of that all Node fetches generate new Node interface objects. It has 22
blocks of tests, where the first 10 are written and the last 12 are not yet
written; those will come in some future release; however, the existing
blocks 5 thru 10 of the present ones are like the missing ones, where the
truly unique ones, 1-4, are done already.
* Corresponding changes in Routine.pm, Details.pod, en.pm: Added new core
class SQL::Routine::Group (Group), objects of which are associated with
Container interface objects and provide convenient means to address a group
of Nodes as a unit, mainly in regards to defining access controls. In
Details.pod, added new main documentation section "ABOUT NODE GROUPS", and
in Routine.pm, updated the "CLASSES IN THIS MODULE" documentation. In the
code of Routine.pm, added the Group class declaration (with 2 default
properties, 'container' and 'member_nsrefs'), which has its own new()
constructor function, but which currently has no other methods (some were
started, but they're commented out); also added new SQL::Routine wrapper
function new_group(). Added documentation for those 2 new constructor
functions in Details.pod. Added 2 new en.pm user text messages for
Group.new(), 'SRT_G_NEW_GROUP_[NO_ARG|BAD]_CONT'.
* Corresponding changes in Routine.pm, Details.pod, en.pm: Added the
beginnings of some new core features, such as shared and exclusive locks on
Nodes, transactions and near-ACID compliance. In Details.pod, added 2 new
sub-sections of FAULT TOLERANCE, "ACID Compliance and Transactions" and
"Access Controls and Reentrancy"; they outline and describe how these
features are intended to be once complete. In the code of Routine.pm, added
new properties to 3 classes; Container interface gained 2, 'explicit_groups'
and 'default_group'; NodeStorage gained 4,
'att_[write|pc_add|lc_add]_blocks' and 'att_mutex'; Group gained 4,
'is_[write|pc_add|lc_add]_block' and 'is_mutex'; none of them have accessor
methods, as of yet or ever. Also updated the Routine.pm code for these 20
methods to add references to the new properties: Container.new(),
Container.new_interface(), Container.delete_node_tree(),
move_before_sibling(), NodeStorage._new(), _delete_node[|_tree](),
_set_node_id(), _[clear|set]_primary_parent_attribute(),
_clear_[literal|enumerated|node_ref]_attribute[|s](),
_set_[literal|enumerated|node_ref]_attribute(), Group.new(). Also added the
4 en.pm user text messages 'SRT_[C|N]_METH_VIOL_WRITE_BLOCKS' and
'SRT_N_METH_VIOL_[P|L]C_ADD_BLOCKS'.
2005-07-16 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.67.
* New code file versions are: Routine.pm v0.67.
* This release is entirely dedicated to refactoring the functions and
methods of Routine.pm (there were zero changes to the property lists and
object lists); all other distribution core files had zero changes, and there
was only a single non-version-check line changed in the test suite. The
public API is exactly the same, save that some thrown exceptions are now
declared to be thrown by different methods than before. The total function
+ method count (public and private) is now 144, up from 121 of the previous
release. All *Storage functions/methods are private, and most * interface
methods are public.
* Split the wrapper functions new_[container|node]() into themselves and
_new_[con|nod]stor(); all 4 belong to the 'Rosetta' superclass. Split both
[Container|Node].new() constructors into themselves and
[Container|Node]Storage._new(). Split each [Container|Node]._ns_to_ni()
private method into itself and [Container|Node]._ns_to_ni_item().
* Renamed Container private method _assert_deferrable_constraints() to
_assert_child_deferrable_constraints().
* Moved these 1 private methods from Container to ContainerStorage:
_assert_child_deferrable_constraints().
* Split these 3 public Container methods each 2-ways into themselves and
_*() private methods, which live in ContainerStorage:
find_child_node_by_surrogate_id(), assert_deferrable_constraints(),
get_all_properties().
* Moved these 20 private methods from Node to NodeStorage:
_delete_node_tree__add_to_candidates(), _get_primary_parent_attribute(),
_clear_primary_parent_attribute(), _set_primary_parent_attribute(),
_get_literal_attribute(), _clear_literal_attribute(),
_set_literal_attribute(), _get_enumerated_attribute(),
_clear_enumerated_attribute(), _set_enumerated_attribute(),
_get_node_ref_attribute(), _clear_node_ref_attribute(),
_set_node_ref_attribute(), _find_node_by_surrogate_id_remotely(),
_find_node_by_surrogate_id_remotely_below_here(),
_find_node_by_surrogate_id_within_layers(),
_find_node_by_surrogate_id_using_path(),
_assert_in_node_deferrable_constraints(),
_assert_parent_ref_scope_deferrable_constraints(),
_assert_child_comp_deferrable_constraints().
* Split these 14 public Node methods each 2-ways into themselves and _*()
private methods, which live in NodeStorage: delete_node(),
delete_node_tree(), get_node_id(), set_node_id(),
clear_literal_attributes(), clear_enumerated_attributes(),
clear_node_ref_attributes(), get_surrogate_id_attribute(),
get_surrogate_id_chain(), find_node_by_surrogate_id(),
find_child_node_by_surrogate_id(), get_relative_surrogate_id(),
assert_deferrable_constraints(), get_all_properties().
* A 'SRT_[C|N]_METH_ASS_READ_ONLY' exception is now only thrown in about
half of the places it used to be (the count was 41 and is now 20); the
removed ones were higher-level functions that always (sans other exceptions)
invoked lower level functions that threw those same exceptions. The effect
of this change is that some of these exceptions will claim to be thrown by
different methods than they used to, under the same user invocation. With
the remaining exception instances, they are also now thrown by the private
part of a method pair, rather than the public part.
* Updated the last test in SRT_10_ReadOnly.t to expect a
'SRT_C_METH_ASS_READ_ONLY' exception rather than a
'SRT_N_METH_ASS_READ_ONLY' exception; this was an exactly 1 letter change.
2005-07-12 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.66.
* New code file versions are: Routine.pm v0.66 and en.pm v0.35.
* Corresponding data dictionary update in NodeTypes.pod, Routine.pm, en.pm:
Added "Distinct Child Groups List" constraint on the 'tools' pseudo-Node
that constrains the 'product_code' attribute of
'data_[storage|link]_product' Nodes. This means that external code can be
confident that every one of those Nodes represents a distinct product
without having to compare their 'product_code' attributes. Also updated
Routine.pm's _assert_child_comp_deferrable_constraints() method to split up
a couple assertions into variants for Node vs pseudo-Node parents (the
pseudo-Node version was omitted before as it was known to not be in use);
correspondingly, these 2 en.pm user text messages were each split in 2:
'SRT_N_ASDC_CH_N_TOO_MANY_SET' and 'SRT_N_ASDC_MUDI_NON_DISTINCT' became
themselves plus '*_PSN' variants.
* Minor Details.pod and Routine.pm update so that any Perl subs which were
referred to as "function/method" etc are now simply "function".
* The private method _normalize_primary_parent_or_node_ref_attribute_value()
was removed; its functionality was added to _set_primary_parent_attribute()
and _set_node_ref_attribute(), which were its only invokers before. While
the change introduces a small amount of duplicated functionality, the system
should be more maintainable overall. No corresponding en.pm changes.
* This release primarily serves to split up the two SQL::Routine::* classes
into 2 tightly bound classes each, ::Container into itself and
::ContainerStorage, and ::Node into itself and ::NodeStorage. With of this
release, the 2 'Storage' classes contain all of the old properties and the 2
suffix-less (interface) classes contain all of the old methods. Each
'Storage' class is an inner class of its corresponding interface class, and
external code can only access the former by way of the latter. Each
interface object has a single Storage object as its sole (almost) property.
For all intents and purposes, the 2 'Storage' classes are the same as the
old 2 classes, in that their own properties only refer to each other and not
to any interface classes. The interface objects are loosely bound, and
multiples of them can now exist for the same Storage objects. The result of
this change is fully backwards compatible but for one thing; you can no
longer reliably do a straight comparison between 2 Container or Node
interface objects to test that they are the same object (eg, like "$foo_node
eq $bar_node"). Instead, you must now invoke the newly added get_self_id()
method on both interface objects you wish to compare; its character string
return value is reliable for this purpose of determing that 2 objects are
the same aka have a Storage object in common. Also added new private method
_ns_to_ni() to each interface class, and they are invoked in 45 places; the
list of other methods and functions is unchanged from before. Except for
adding new_interface() and get_self_id() to the method documentation in
Details.pod and adding a new RELATING INTERFACE AND STORAGE CLASSES section
there, the entirety of changes for this release item are in Rosetta.pm
itself, and almost entirely in the code. A side benefit of this update is
you can once again keep a SQL::Routine model in existence by holding a
reference to a Container or a Node, but this time no strong circular
references are in use.
* Added some TODO file items.
2005-06-17 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.65.
* New code file versions are: Routine.pm v0.65 and en.pm v0.34.
* Updated external dependencies on Locale::KeyedText to v1.05.
* Updated all files in this distribution that display the physical address
of the Free Software Foundation, Inc. so it uses their current address of
"51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA" rather than their
previous address of "59 Temple Place, Suite 330, Boston, MA 02111-1307 USA";
this affects every COPYRIGHT AND LICENSE statement in the core files and the
'ReadMe' file, plus any 'GPL' or 'LGPL' license file.
* Updated the INSTALL file so that it is more generic, and all of my
distributions now use an identical copy of it.
* Updated the main heading in the ReadMe file to state the distribution's
name rather than a terse description of it.
* Updated Routine.pm and Details.pod to split the Container property
'def_con_tested' into itself and 'edit_count'; where the old property was a
boolean, the two new ones are both integers. The new 'edit_count' property
exists to track that changes were made to a Container's Node set; it is
incremented each time a Node is edited, added, or deleted; its actual value
isn't important, only that two arbitrary samplings of it will be different
if an edit happened between them. This feature is designed to assist
external code that caches information derived from a SQL::Routine model,
such as generated SQL strings or Perl closures, so that it can easily tell
when the cache may have become stale (leading to a cache flush). The new
'def_con_tested' stores a sampled copy of 'edit_count'. Within Routine.pm
itself, every line of code that used to set 'def_con_tested' to false now
increments 'edit_count' instead; the single code line that used to set
'def_con_tested' to true now copies the value of 'edit_count' to it instead;
the 2 lines of code that used to read the old 'def_con_tested' value now
compare its value to 'edit_count' instead. Added new read-accessor method
for 'edit_count' named get_edit_count(). Updated the POD documentation for
deferrable_constraints_are_tested() and
Container.assert_deferrable_constraints(); but their behaviour is unchanged.
* In Details.pod, updated get_next_free_node_id() POD; no behaviour changes.
* Updated data dictionary in Routine.pm and NodeTypes.pod and Language.pod:
Now 'view' Nodes can no longer be primary-children of 'routine' Nodes;
instead, they can now be primary-children of either 'routine_var' or
'routine_stmt' Nodes; do the first with all cursor definitions (only where
the cursor is to be declared, not for any other routine_var that may hold it
later), and do the second usually with SELECT or DML statement definitions.
A result of this change is that the Node tree for a routine now looks more
natural when compared to the routine string-SQL that would be generated from
it; moreover, one no longer has to concern themselves to ensure that the
relevant 'view' Node si_names are distinct relative to all variable etc
names in the routine, which was an artificial constraint to begin with.
* Updated data dictionary in Routine.pm and NodeTypes.pod and Language.pod
(with mention also in Details.pod): Added new Node type 'external_cursor'
which lives under 'elements'; it contains no attributes of importance and
exists simply as a common point to be referred to by all routines that share
a cursor (one of those pointing to it will hold the cursor definition); also
renamed all 3 'curs_view' attributes (in 'routine_[context|arg|var]' Nodes)
to 'curs_ext', and they all point to an 'external_cursor' Node now instead
of a 'view' Node.
* Updated data dictionary in Routine.pm and NodeTypes.pod: Added 2 new
'routine' Node attributes, 'return_conn_link' and 'return_curs_ext', plus
related constraints; these are used in the same way as the 'conn_link' and
'curs_ext' attributes of the 'routine_context' and 'routine_arg' Nodes.
* Updated the SYNOPSIS documentation of Routine.pm to adjust for all the
abovementioned data dictionary updates. Likewise updated all 4 core
t_SRT_*.pm test files to remain compatible with the changes. Mostly this
involved rearranging the order of Nodes; but in Terse|Abstract there were
also 4 new external_cursor Nodes added plus corresponding 'routine' attrs.
* The documentation file API_C.pod no longer exists as its own entity; all
of its distinct content was merged into the file Details.pod, occupying a
new main documentation section at its end named OLDER DOCUMENTATION. Since
any possibility of making a C version won't be realized for a long time,
with a Parrot-based version more imminent, I've decided to de-emphasize (but
not discard) the C-specific details; the lack of their own file makes these
details a lot less visible.
* Added TODO file item about adding better support for arguments to cursors
that are shared between multiple routines.
* The core files en.pm and EnumTypes.pod had zero changes aside from the FSF
address update.
2005-06-10 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.64.
* New code file versions are: Routine.pm v0.64 and en.pm v0.33.
* Updated Routine.pm and Details.pod to change the semantics of a
Container's "deferrable constraints are tested" property such that it is
always set false by default on a new Container (rather than true) since a
Container completely devoid of Nodes may violate a deferrable constraint;
aside from the initial valuation change made in Container.new() and related
documentation updates by the property's declaration and the
deferrable_constraints_are_tested() method, there were no other changes
concerning this property.
* Updated Routine.pm, Details.pod, en.pm: Added new Routine.pm Node public
method delete_node_tree(), which will remove a Node plus all of its direct
descendant Nodes. This method is mainly intended for use by long-lived
applications that continuously generate database commands at run-time and
only use each one once, such as a naive interactive SQL shell (or a naive
DBI module emulator, which ties this function's use to a $sth destruction),
so to save on memory use. Also added new en.pm user text message
'SRT_N_DEL_NODE_TREE_HAS_EXT_CHILD' which is used by it. Also added new
private helper method _delete_node_tree__add_to_candidates(). Also added
new Container method delete_node_tree() which deletes all Nodes in a
Container. Also made an execution speed optimization to delete_node(), and
updated its 'SRT_N_DEL_NODE_HAS_CHILD' message to give more details.
* Corresponding updates in NodeTypes.pm, Routine.pm to correct a small data
dictionary design flaw concerning the "Ancestor Attribute Correlation List",
which affects view_expr and routine_expr Nodes. The flaw is summarized that
certain steps while navigating a Node tree are assumed to always result in
the observer sitting on a Node of a specific type, whereas they may
sometimes be sitting on a different type instead; any subsequent steps would
then be invalid. The fix involves making such subsequent steps conditional
on the current Node's type. Updated data dictionary entries for 5 view_expr
(3) and routine_expr (2) Node attributes to be conditional. Added a few
code lines to the _find_node_by_surrogate_id_using_path() method which
implements the added conditional logic.
* Corresponding updates in NodeTypes.pm, Routine.pm, en.pm to add a new
"Related Parent Enumerated Attributes List" deferrable constraint. Its main
purpose is to assert that any arguments given to invocations of standard
routines have correct names; previously, an argument name that was valid for
any single standard routine could be passed to any other standard routine
without being flagged; now, only argument names specifically for the routine
being invoked are allowed. Updated the data dictionary entries for
view_expr and routine_expr Nodes to use this new constraint. Added new
global constant hash %P_C_REL_ENUMS in Routine.pm, which specifies the valid
relations between pairs of enumerated type values. Also added a new code
block to the _assert_parent_ref_scope_deferrable_constraints() method which
implements the new constraint assertion; also added 4 related new
'SRT_N_ASDC_REL_ENUM_[BAD_P_NTYPE|NO_P|P_NEVER_P|P_C_NOT_REL]' user text
messages to en.pm.
* Corresponding updates in NodeTypes.pm, Routine.pm, en.pm to add a new
"Mandatory Related Child Enumerated Attributes List" deferrable constraint.
Its main purpose is to assert that all invocations of standard routines have
with them a complete set of standard routine arguments (and context), and
that no extra child Nodes exist for the invocation Node that don't provide
args/context; previously, missing arguments were not flagged. Updated the
data dictionary entries for view_expr and routine_stmt and routine_expr
Nodes to use this new constraint. Added new global constant hash
%OPT_P_C_REL_ENUMS in Routine.pm, which specifies which arguments in
%P_C_REL_ENUMS are optional. Also added a new code block to the
_assert_child_comp_deferrable_constraints() method which implements the new
constraint assertion; also added 3 related new
'SRT_N_ASDC_MA_REL_ENUM_[TOO_MANY_SET|ZERO_SET|MISSING_VALUES]' user text
messages to en.pm. Also updated _build_node_is_child_or_not() to catch the
new 'SRT_N_ASDC_MA_REL_ENUM_MISSING_VALUES' exceptions when auto-asserting
deferrable constraints.
* Updated all 4 main t_SRT_*.pm files to perform bug fixes for input errors
that were flagged by the new "Mandatory ..." constraint (no fixes were made
concerning the other new or corrected constraints, as the tests already
obeyed them); also updated the corresponding expected XML strings. In
t_SRT_Synopsis.pm, modified 2 Node definitions; also made identical
corrections in the SYNOPSIS of Routine.pm, which that test mirrors. In
t_SRT_Verbose.pm, modified 4 Node definitions. In each of t_SRT_Terse.pm
and t_SRT_Abstract.pm, added 9 Node definitions, deleted 1 other, and
modified 10 others; also did some code reformatting.
* A few trivial updates to a few Node.assert_deferrable_constraints()
methods in Routine.pm. Changed any error message arguments that interpolate
arrays into strings so they pass an array ref instead (which is later
interpolated in a different way). Corrected any code comments to refer to
NodeTypes.pod rather than Language.pod. Updated some other code comments.
* Zero changes to these core files: Language.pod, EnumTypes.pod, API_C.pod.
2005-06-01 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.63.
* New code file versions are: Routine.pm v0.63 and en.pm v0.32.
* In summary, this release primarily serves to improve the ease of use and
reliability of the build_*() methods, such that their allowed argument
options are now more consistent and predictable. The allowed argument
formats are now more terse and user-friendly than before. The
get_all_properties() methods will now return the more terse format as well.
Finally, the number of static error messages has been greatly reduced,
replaced with a much smaller number of dynamic ones, so the total is easier
to work with.
* Refactored a lot of user text messages in en.pm by combining 46
method-specific messages into 9 method-generic messages taking more
variables (an 80% compression rate); in the process, en.pm became 22%
(7K/33K) smaller. Routine.pm also had its corresponding methods updated to
use the replacement messages. The 1 'SRT_C_METH_ARG_UNDEF' replaced these
2: 'SRT_C_FIND_[NODE_BY_ID_NO_ARG_ID|CH_ND_BY_SID_NO_ARG_VAL]'. The 1
'SRT_N_METH_ARG_UNDEF' replaced these 24:
'SRT_N_SET_[NODE_ID_NO_ARGS|PP_AT_NO_ARG_VAL]',
'SRT_N_[GET|CLEAR|SET]_[LIT|ENUM|NREF]_AT_NO_ARGS',
'SRT_N_SET_[LIT|ENUM|NREF]_ATS_NO_ARGS', 'SRT_N_SET_SI_AT_NO_ARGS',
'SRT_N_[GET|CLEAR|SET]_AT_NO_ARGS', 'SRT_N_SET_ATS_NO_ARGS',
'SRT_N_MOVE_PRE_SIB_NO_S_ARG', 'SRT_N_ADD_CH_NODE_NO_ARGS',
'SRT_N_FIND_[ND_BY_SID_NO_ARGS|CH_ND_BY_SID_NO_ARG_VAL]'
'SRT_N_GET_REL_SID_NO_ARGS'. The 1 'SRT_N_METH_ARG_NO_HASH' replaced these
4: 'SRT_N_SET_[LIT|ENUM|NREF]_ATS_BAD_ARGS', 'SRT_N_SET_ATS_BAD_ARGS'. The
1 'SRT_N_METH_ARG_NO_NODE' replaced these 3:
'SRT_N_MOVE_PRE_SIB_BAD_[S|P]_ARG', 'SRT_N_ADD_CH_NODE_BAD_ARG'. The 1
'SRT_N_METH_NO_PP_AT' replaced these 3:
'SRT_N_[GET|CLEAR|SET]_PP_AT_NO_PP_AT'. The 2
'SRT_N_METH_ARG_NO_[LIT|ENUM]_AT_NM' replaced these 6:
'SRT_N_[GET|CLEAR|SET]_[LIT|ENUM]_AT_INVAL_NM'. The 1
'SRT_N_METH_ARG_NO_NREF_AT_NM' replaced these 5:
'SRT_N_[GET|CLEAR|SET]_NREF_AT_INVAL_NM',
'SRT_N_[FIND_ND_BY|GET_REL]_SID_INVAL_NM'. The 1 'SRT_N_METH_ARG_NO_AT_NM'
replaced these 3: 'SRT_N_[GET|CLEAR|SET]_AT_INVAL_NM'.
* In Routine.pm, refactored Container.build_child_node_tree() to inline what
build_child_node() does rather than invoking it; corresponding to this,
en.pm saw the user text message 'SRT_C_BUILD_CH_ND_NO_PSND' split into
itself and 'SRT_C_BUILD_CH_ND_TR_NO_PSND'. Node.build_child_node_tree() was
likewise refactored to inline build_child_node(), and Node.build_node() to
inline Container.build_node(), but for both there were no en.pm messages
affected by this.
* Minor _build_node_is_child_or_not() correction to invoke user text message
'SRT_N_NEW_NODE_NO_ARG_TYPE' rather than 'SRT_N_NEW_NODE_NO_ARGS'.
* Updated Routine.pm and Details.pod to rename all LIST arguments to
CHILDREN, which affects these methods: add_child_nodes(),
build_child_node[|_tree]s(), build_container(). Also renamed
add_child_node() argument from NEW_CHILD to CHILD.
* Corresponding updates in Routine.pm, Details.pod, en.pm that remove global
support for the alternate method input format of "named arguments in a
single hash-ref argument"; all of the [Container|Node].build_*() methods
were affected, with these 6 being the ones edited: build_node(),
build_child_node(), build_child_node_tree().
* Updated Routine.pm, Details.pod, en.pm so that it is now mandatory for any
defined CHILDREN arguments to be array refs where each element is an array
ref having the details for one child or child tree (mapping to
build[_child]_node() positional arguments). It was previously possible to
pass just the details for a single child Node in place of the first array
ref, but that is no longer valid. A further change is that, for the
build_child_nod[e_tree]s() methods, the CHILDREN argument is now enforced
mandatory; previously those methods would no-op if it was undefined, which
permitted other methods to invoke them regardless of the passed argument
being valued. Updated these methods to only invoke the aforementioned if
they have defined CHILDREN to pass: build_child_node_tree(),
build_container(). Added these 6 new method-generic user text messages to
en.pm, which are used by the newly updated 4 build_*s() methods:
'SRT_[C|N]_METH_ARG_NO_ARY', 'SRT_[C|N]_METH_ARG_ARY_ELEM_[UNDEF|NO_ARY]'.
* Updated the populate_model() functions in t_SRT_[Terse|Abstract].pm to
invoke the build_*() methods with the same argument formats that
t_SRT_Synopsis.pm has already been using, since they are now the only valid
build_*() input formats following the above-mentioned Routine.pm changes;
the expected XML output strings are identical to before. Following the
changes, each of those test suite files is 4K smaller than before.
* Updated Routine.pm and Details.pod to alter the data format that
get_all_properties() and get_all_properties_as_perl_str() return, such that
each Node is represented by a 3-element array-ref, rather than a 3-element
hash-ref; the get_all_properties_as_xml_str() method's return values did not
change at all, however. To implement this, the only methods that actually
saw a code change were: [Container|Node].get_all_properties(),
_serialize_as_perl(), _serialize_as_xml().
* Updated the SRT_50_Cloning.t test script so it invokes build_container()
with an array ref wrapped around the get_all_properties() return value.
* There were zero data dictionary updates in this release, and zero changes
to the Verbose and Synopsis test suites, and no other documentation changes.
2005-05-13 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.62.
* New code file versions are: Routine.pm v0.62 and en.pm v0.31.
* Updated external dependencies on Locale::KeyedText to v1.04.
* Updated every instance of the COPYRIGHT AND LICENSE documentation in this
distribution to correct the declared date range in which the core
SQL::Routine files were created. The new dates are 2002 thru 2005, which is
when I actually committed documentation describing or code implementing what
became the SQL::Routine core. The old dates were 1999 thru 2005, which goes
back to the time that I started conceiving a larger project which the
aforementioned module started out being related to; however, those earlier
years did not contain any work on what specifically became the
aforementioned module, so they are now excluded.
* Updated every instance of the COPYRIGHT AND LICENSE documentation in this
distribution so that you may now choose between licensing it under the GPL
version 2 or any later GPL version, rather than only under version 2.
* Renamed the "LICENSE" file to "GPL", and correspondingly updated all
references to it in this distribution, most of them being in COPYRIGHT AND
LICENSE statements. This change was made to reduce user confusion in
situations where files from this distribution may be aggregated with other
files that have different licenses; the old name was too generic.
* Updated every instance of the COPYRIGHT AND LICENSE documentation in this
distribution to remove the parenthetical references to "libSQLRT"; there
are no references left to a possible C library's name except in API_C.pod.
* Updated SRT_00_Compile.t to add 14 tests, bringing the total to 22; the
new tests are all can_ok(), one for each of the test suite modules' methods.
* Updated data dictionary in Routine.pm, EnumTypes.pm, NodeTypes.pm: Updated
standard-routines CURSOR_FETCH() and SELECT() to add a new INTO argument to
each, which takes a ROW or RW_ARY variable to put the query result in; added
corresponding routine_expr Node attribute 'query_dest' to use when the
Node's call_sroutine_arg is 'INTO'. Removed 'view' Node attribute
'set_p_routine_item', which is how the 'into' information used to be stored
in a model. This change was made so that cursors can be used more easily,
such as when you want different cursor-fetch calls to put their results in
different variables, or you want to open a cursor in one routine while
fetching from it in another. Updated the Routine.pm SYNOPSIS code/XML and
the test file t_SRT_Verbose.pm to remain compatible with the changes.
* Added 2 new test suite files, t/lib/t_SRT_Synopsis.pm and
t/SRT_43_Synopsis.t, which mirror the Routine.pm SYNOPSIS code/XML and makes
sure it works; this adds 6 numbered tests. Updated SRT_00_Compile.t to test
that the new test module compiles; this adds 5 numbered tests.
* Updated data dictionary in Routine.pm, NodeTypes.pm: Updated the
'assign_dest' attribute of routine_stmt Nodes so that 'routine_context' is
no longer a valid Node type to refer to; it does not make sense to assign to
a routine's context, which is analagous to $self.
* Small documentation updates in EnumTypes.pm concerning standard-routines.
* Updated every t/SRT_*.t file except 00_Compile to remove its explicit "use
SQL::Routine::L::en;" line, which is redundant with what LKT does.
2005-04-27 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.61.
* New code file versions are: Routine.pm v0.61 and en.pm v0.30.
* This release saw a large reorganization of the test suite. The file
t/SQL_Routine.t was split-up into these 6 parts (numbers in parens are the
portion of the original 35 numbered tests that each part got):
t/lib/t_SRT_Util.pm (zero), t/SRT_11_Circular.t (2), t/SRT_40_Verbose.t (6),
t/SRT_41_Terse.t (6), t/SRT_42_Abstract.t (6), t/SRT_50_Cloning.t (15). The
test utility module t_SRT_Util.pm contains the message() and
error_to_string() functions used by most of the other test scripts; the
message() function also received a minor update such that its output is
prefixed with '#' rather than '--', like Test::More does for comments.
* Added new test script t/SRT_00_Compile.t which contains 8 brand new
numbered tests; 6 of these use Test::More's use_ok() function to cleanly
test that the 2 core modules and 4 test modules will load without errors; 2
of these test, with cmp_ok(), that the 2 core modules are of the correct
versions. This new script runs prior to all of the others. Before this
change, the original test script would simply die without any ok/not ok.
* Updated Routine.pm and Details.pod and en.pm: Added new boolean Container
property, "is read only", plus an accessor method named is_read_only(); any
attempt to modify or add/remove the Container's Nodes or most of its other
properties will throw an exception when the property is true. Updated 8
Container methods plus 31 Node methods to perform the assertions; also added
2 related user text messages to en.pm, 'SRT_[C|N]_METH_ASS_READ_ONLY', which
are shared by all the methods.
* Added new test script t/SRT_10_ReadOnly.t, which has 17 numbered tests,
to test the new read-only functionality.
* Updated the FAULT TOLERANCE AND MULTI-THREADING SUPPORT main documentation
section of Details.pod, concerning types of input validation.
* Updated the MANIFEST for the 8 added files and 1 removed file. Also
updated the INSTALL file.
* There were zero changes to all 4 of the other /lib/SQL/Routine/* files and
all 3 of the other /t/lib/t_SRT_*.pm files.
2005-04-25 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.60.
* New code file versions are: Routine.pm v0.60 and en.pm v0.29.
* The primary purpose of this release is the removal of the long-standing
feature where a SQL::Routine Node is valid to exist either on its own or
inside a Container; these 2 states of a Node were referred to as "alone" and
"well known" respectively; as of this release, there is only a single valid
state, which corresponds to the old "well known". One result of this change
is that the SQL::Routine code can be greatly simplified, and it probably
executes faster too. Note that this constitutes an API change, but the only
external code that would be affected is that which directly invokes these
low-level functions and methods: new_node(), Node.new(), clear_node_id(),
put_in_container(), take_from_container(), build_lonely_node(). Any code
not invoking these, such as because they use the other high-level build*()
functions instead (including all of the SYNOPSIS code), will not break.
* All changes in this release are restricted to just the following 3 files
unless otherwise explicitly mentioned: Routine.pm and Details.pod and en.pm.
* Reversed the change in release 0.59 that removed the Node.delete_node()
method; its concept is useful after all.
* It is now the case that every Node's "container" and "node id" properties
are always set and never empty, from the moment the Node object is created
and up to its destruction. Any branch conditionals that tested those
properties for definedness were reduced to just the result of the 'yes'
branch; this includes the removal of many throwable exceptions.
* In Details.pod, deleted the entire NODE EVOLUTION STATES main
documentation section, since there is no longer a multiplicity of Node
states as it describes. Many other small bits of POD documentation and code
comments were updated to remove any references to 'alone' or 'well known';
any documentation that described an either/or scenario pair was reduced to
one always-true scenario.
* Changed the declaration order of a Node's "container" property so it
appears just above "node type"; also moved the method get_container()
upwards to just above get_node_type().
* Merged the method put_in_container(), plus a copy of set_node_id(), into
Node.new(); the post-merge new() now creates a Node that lives in a
Container and has a set Node Id; it does not create links with any other
Nodes, but only the pseudo-Node parent; it now takes 3 arguments: a
Container, a Node Type, and a Node Id. The merged method has 7 associated
user text messages in en.pm, down from the 9 its 2 main constituents had.
Updated the new_node() wrapper function to match the new 3 arg signature.
* Merged the method take_from_container() into delete_node(); the post-merge
delete_node() unlinks the invocant Node from everything else and then gets
it garbage collected. The merged method has 1 user text message in en.pm,
down from the 2 its constituents had. Updated the method
Container.build_child_node() to invoke it rather than take_from_container().
* Deleted method clear_node_id(), along with its single en.pm message. Also
removed its invocation from clear_attributes(), and updated the invocations
in clear_surrogate_id_attribute() and clear_attribute() to throw an
exception where they had used to invoke it; also added 2 new '*_MAND_NID'
en.pm messages to go with those last 2 methods.
* Updated _normalize_primary_parent_or_node_ref_attribute_value() and en.pm
to remove 2 assertions and 4 associated exception messages that will no
longer be triggered, SRT_N_SET_[PP|NREF]_AT_[ONE_CONT|MISS_NID].
* Updated these 4 methods so that they always run their payload rather than
only doing so if the nref-attr value (that was or to be) is an object
reference: _[clear|set]_[primary_parent|node_ref]_attribute().
* Removed function build_lonely_node(). Then merged private function
_build_node_normalize_attrs() into its only remaining invoker, private
method _build_node_is_child_or_not().
* Updated private method _assert_in_node_deferrable_constraints() so that it
no longer tests for the Node Id being defined; that is now redundant. Also,
a related piece of documentation in NodeTypes.pod was updated, to say that
the 'id' attribute now has a constantly applied mandatory constraint.
* Fixed a bug in _build_node_is_child_or_not() where it only trapped
SRT_N_ASDC_CH_N_TOO_FEW_SET exceptions and not the similar
SRT_N_ASDC_CH_N_TOO_FEW_SET_PSN exceptions.
* Fixed a bug in _assert_child_comp_deferrable_constraints() where the
SRT_N_ASDC_CH_N_TOO_FEW_SET_PSN exception was missing a variable value.
* Updated _build_node_is_child_or_not() so that it always sets the new
Node's surrogate id attribute prior to any other non-id/pp attribute; this
ensures that if any other attribute settings fail, the error messages are
more user friendly.
* Updated the following 7 methods to not assert their invocant Node is in a
Container, and removed their corresponding '*_NOT_IN_CONT' en.pm messages:
move_before_sibling(), get_surrogate_id_chain(),
find[|_child]_node_by_surrogate_id(), get_relative_surrogate_id(),
build[|_child]_node(). Also updated all 15 Node methods that conditionally
(based on the Container's presence) set the 'deferrable constraints tested'
property to false so they don't test.
* Merged _get_surrogate_id_chain() into get_surrogate_id_chain().
* Merged each _get_all_properties() into its partner get_all_properties().
* Updated the UNDERSTANDING NODE ATTRIBUTE ... section of NodeTypes.pod to
remove references to Node states and to say that the mandatory constraint
Node Ids is now constantly enforced rather than being deferrable.
* In Routine.pm, deleted the second CAVEATS paragraph, concerning Nodes that
outlive their host Container; deleted a similar section in the TODO file.
* Updated the sqlrt_dt_node struct definition in API_C.pod so that the
'pp_nref' and 'at_nrefs' members are plain refs/pointers and not unions of
pointers and unsigned integers. This corresponds to the Routine.pm changes.
* Small test suite update: Updated the 2 make_a_[|child_]node small utility
functions in t_SRT_Verbose.pm; folded the set_node_id() and
put_in_container() invocations into the new_node() invocation. No changes
in SQL_Routine.t besides the usual upping of the 'used' module versions.
* These core files had exactly zero changes: EnumTypes.pod, Language.pod,
t_SRT_Terse.pm, t_SRT_Abstract.pm.
* Following all of the above changes, the disk space for all 7 core /lib/SQL
files in release 0.60 is 430K, which is 16K (or 4%) less than the 446K used
by the same files in release 0.59. Specifically with the executable
function+method code (including spaces and dividers and f+m code comments)
in Routine.pm, 0.60 is 82K, which is 7K (or 9%) less than the 89K for 0.59;
by line count, 0.60 is 2092 f+m code lines, which is 162 (or 7%) lines less
than the 2254 lines of 0.59.
* For speed comparisions between 0.60 and 0.59, based on 100 iterations of
most of the test suite; as we can see, 0.60 performs a bit faster by up to
15% or so:
Version 0.60:
Benchmark: timing 100 iterations of A_AssDefCon, A_ConvertToXML_NID, A_ConvertToXML_SID, A_MakeAndBreak, T_AssDefCon, T_ConvertToXML_NID, T_ConvertToXML_SID, T_MakeAndBreak, V_AssDefCon, V_ConvertToXML_NID, V_ConvertToXML_SID, V_MakeAndBreak...
A_AssDefCon: 13 wallclock secs (12.72 usr + 0.00 sys = 12.72 CPU) @ 7.86/s (n=100)
A_ConvertToXML_NID: 2 wallclock secs ( 2.38 usr + 0.02 sys = 2.40 CPU) @ 41.67/s (n=100)
A_ConvertToXML_SID: 6 wallclock secs ( 6.20 usr + 0.04 sys = 6.24 CPU) @ 16.03/s (n=100)
A_MakeAndBreak: 14 wallclock secs (13.35 usr + 0.01 sys = 13.36 CPU) @ 7.49/s (n=100)
T_AssDefCon: 12 wallclock secs (12.81 usr + 0.01 sys = 12.82 CPU) @ 7.80/s (n=100)
T_ConvertToXML_NID: 2 wallclock secs ( 2.51 usr + 0.02 sys = 2.53 CPU) @ 39.53/s (n=100)
T_ConvertToXML_SID: 7 wallclock secs ( 6.41 usr + 0.03 sys = 6.44 CPU) @ 15.53/s (n=100)
T_MakeAndBreak: 7 wallclock secs ( 7.88 usr + 0.00 sys = 7.88 CPU) @ 12.69/s (n=100)
V_AssDefCon: 6 wallclock secs ( 5.66 usr + 0.01 sys = 5.67 CPU) @ 17.64/s (n=100)
V_ConvertToXML_NID: 1 wallclock secs ( 1.33 usr + 0.00 sys = 1.33 CPU) @ 75.19/s (n=100)
V_ConvertToXML_SID: 3 wallclock secs ( 2.97 usr + 0.00 sys = 2.97 CPU) @ 33.67/s (n=100)
V_MakeAndBreak: 3 wallclock secs ( 2.95 usr + 0.00 sys = 2.95 CPU) @ 33.90/s (n=100)
Version 0.59
Benchmark: timing 100 iterations of A_AssDefCon, A_ConvertToXML_NID, A_ConvertToXML_SID, A_MakeAndBreak, T_AssDefCon, T_ConvertToXML_NID, T_ConvertToXML_SID, T_MakeAndBreak, V_AssDefCon, V_ConvertToXML_NID, V_ConvertToXML_SID, V_MakeAndBreak...
A_AssDefCon: 12 wallclock secs (12.81 usr + 0.00 sys = 12.81 CPU) @ 7.81/s (n=100)
A_ConvertToXML_NID: 2 wallclock secs ( 2.38 usr + 0.01 sys = 2.39 CPU) @ 41.84/s (n=100)
A_ConvertToXML_SID: 6 wallclock secs ( 6.31 usr + 0.02 sys = 6.33 CPU) @ 15.80/s (n=100)
A_MakeAndBreak: 15 wallclock secs (14.66 usr + 0.00 sys = 14.66 CPU) @ 6.82/s (n=100)
T_AssDefCon: 13 wallclock secs (13.03 usr + 0.00 sys = 13.03 CPU) @ 7.67/s (n=100)
T_ConvertToXML_NID: 3 wallclock secs ( 2.54 usr + 0.01 sys = 2.55 CPU) @ 39.22/s (n=100)
T_ConvertToXML_SID: 6 wallclock secs ( 6.45 usr + 0.03 sys = 6.48 CPU) @ 15.43/s (n=100)
T_MakeAndBreak: 9 wallclock secs ( 9.01 usr + 0.00 sys = 9.01 CPU) @ 11.10/s (n=100)
V_AssDefCon: 6 wallclock secs ( 5.70 usr + 0.00 sys = 5.70 CPU) @ 17.54/s (n=100)
V_ConvertToXML_NID: 1 wallclock secs ( 1.28 usr + 0.01 sys = 1.29 CPU) @ 77.52/s (n=100)
V_ConvertToXML_SID: 3 wallclock secs ( 3.02 usr + 0.00 sys = 3.02 CPU) @ 33.11/s (n=100)
V_MakeAndBreak: 4 wallclock secs ( 3.53 usr + 0.01 sys = 3.54 CPU) @ 28.25/s (n=100)
The RAM usage was roughly the same for both, or slightly less for 0.60.
2005-04-20 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.59.
* New code file versions are: Routine.pm v0.59 and en.pm v0.28.
* Added new external dependency on Scalar::Util v1.11 (bundled with Perl
5.8.x); SQL::Routine now uses its 'weak refs' feature internally at
strategic places so that you no longer need to explicitly destroy Containers
when you are done with them. The circular references between SRT Nodes and
their Container no longer prevent Perl's standard garbage collection from
working when expected, which is when all external references to a Container
go away. This change makes SQL::Routine much more DWIM and easier to use.
* Updated Routine.pm and Details.pod: Removed the Container.destroy()
method. Removed the CAVEATS documentation concerning it. Updated
SQL_Routine.t to remove all destroy() invocations, which also reduced the
test count by 6 to 35. Removed the corresponding TODO file item.
* Updated Routine.pm: Updated the definitions of 3 Node properties to
indicate that they now store weak Perl refs rather than strong ones:
'pp_nref', 'at_nrefs', 'container'. Those 3 properties are all 'upwards' or
'parent pointing' references; each has one or more corresponding 'child
pointing' properties which remain strong refs. As such, every conceptual
circular reference now has one strong and one weak member. A total of 5
Scalar::Util::weaken() invocations are now used (up from zero): 1 is in
Node._set_primary_parent_attribute(); 1 is in Node._set_literal_attribute();
3 are in Node.put_in_container(), one per property.
* Updated Routine.pm and Details.pod and en.pm: Removed the
Node.delete_node() method, which was a no-op routine and never invoked; a
SRT Node has always been garbage collected by Perl as is normal. This
method had existed as a place-holder for an eventual C implementation that
actually did something, but since a C version is very far away, if it ever
comes, this has been removed meanwhile. Also removed the corresponding
en.pm user text message 'SRT_N_DEL_NODE_IN_CONT'.
* Updated both CREDITS copies regarding Stevan Little's contribution, to add
that he introduced me to Scalar::Util::weaken() last month.
* Added new CAVEATS item and TODO file item concerning external refs to
Nodes whose host Container was garbage collected.
* Renamed this file to 'Changes' from 'ChangeLog'; the new name is more
descriptive to its free-form structure, and more consistent with CPAN (which
uses both names but 'Changes' seems to be more common these days). Also
truncated this Changes file to remove all detail entries for releases v0.28
thru v0.58; in their place was added a change summary for releases v0.01
thru v0.58 that mainly just said what the names and versions of all the
predecessor distributions and modules are.
2005-04-03 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.58, containing SQL::Routine 0.58, was released on
CPAN.
2005-03-06 Darren Duncan <perl@DarrenDuncan.net>
Record update in the PAUSE modules database:
statd: [a] was [c]
The resulting entry will be:
SQL::
::Routine adpOg Specify all database tasks with SQL routines DUNCAND
2004-11-05 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.48, containing SQL::Routine 0.48, was released on
CPAN. This is the first release of SQL::Routine following its merger with
SQL::Routine::SkipID.
2004-11-02 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine-Castaways Release 0.33, containing SQL::Routine::SkipID 0.33,
was released on CPAN; this is the last release of the SQL-Routine-Castaways
distribution. This is the last release of SQL::Routine::SkipID prior to its
merger with SQL::Routine.
2004-11-01 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.47, containing SQL::Routine 0.47, was released on
CPAN. This is the last release of SQL::Routine prior to its merger with
SQL::Routine::SkipID. This is the first release of SQL::Routine following
its merger with SQL::Routine::ByTree.
2004-10-29 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.46, containing SQL::Routine 0.46, was released on
CPAN. This is the last release of SQL::Routine prior to its merger with
SQL::Routine::ByTree.
2004-10-04 Darren Duncan <perl@DarrenDuncan.net>
* SQL-Routine Release 0.43, containing SQL::Routine 0.43, was released on
CPAN; this is the first release of the SQL-Routine distribution. This is
the first release of any distribution to contain SQL::Routine, which was
renamed from SQL::SyntaxModel. This is the first release of this module
following its official registration on the Perl 5 Module List.
* SQL-Routine-Castaways Release 0.31, containing SQL::Routine::ByTree 0.31
and SQL::Routine::SkipID 0.31, was released on CPAN; this is the first
release of the SQL-Routine-Castaways distribution. This is the last release
of SQL::Routine::ByTree prior to its merger with SQL::Routine. This is the
first release of any distribution to contain SQL::Routine::ByTree or
SQL::Routine::SkipID; they were renamed from SQL::SyntaxModel::ByTree and
SQL::SyntaxModel::SkipID respectively.
2004-10-04 Darren Duncan <perl@DarrenDuncan.net>
The next version of the Module List will list the following module:
modid: SQL::Routine
DSLIP: cdpOg
description: Specify all database tasks with SQL routines
userid: DUNCAND (Darren Duncan)
chapterid: 11 (String_Lang_Text_Proc)
enteredby: BDFOY (brian d foy)
enteredon: Mon Oct 4 20:04:46 2004 GMT
The resulting entry will be:
SQL::
::Routine cdpOg Specify all database tasks with SQL routines DUNCAND
2004-09-13 Darren Duncan <perl@DarrenDuncan.net>
* SQL-SyntaxModel Release 0.42, containing SQL::SyntaxModel 0.42, was
released on CPAN; this is the last release of the SQL-SyntaxModel
distribution. This is the last release of any distribution to contain
SQL::SyntaxModel. This is the last release of this module prior to its
official registration on the Perl 5 Module List.
2004-09-01 Darren Duncan <perl@DarrenDuncan.net>
* SQL-SyntaxModel-ByTree Release 0.30, containing SQL::SyntaxModel::ByTree
0.30, was released on CPAN; this is the last release of the
SQL-SyntaxModel-ByTree distribution. This is the last release of any
distribution to contain SQL::SyntaxModel::ByTree. This is the first release
for this distribution and module where they both explicitly have matching
version numbers.
* SQL-SyntaxModel-SkipID Release 0.30, containing SQL::SyntaxModel::SkipID
0.30, was released on CPAN; this is the last release of the
SQL-SyntaxModel-SkipID distribution. This is the last release of any
distribution to contain SQL::SyntaxModel::SkipID. This is the first release
for this distribution and module where they both explicitly have matching
version numbers.
2004-08-16 Darren Duncan <perl@DarrenDuncan.net>
* SQL-SyntaxModel Release 0.38, containing SQL::SyntaxModel 0.38, was
released on CPAN. This is the first release for this distribution and
module where they both explicitly have matching version numbers.
2004-08-06 Darren Duncan <perl@DarrenDuncan.net>
* SQL-SyntaxModel Release 0.37, containing SQL::SyntaxModel 0.24, was
released on CPAN. This is the last release for this distribution and
module where they both do not explicitly have matching version numbers.
2004-06-29 Darren Duncan <perl@DarrenDuncan.net>
* SQL-SyntaxModel-ByTree Release 0.29, containing SQL::SyntaxModel::ByTree
0.14, was released on CPAN. This is the last release for this distribution
and module where they both do not explicitly have matching version numbers.
* SQL-SyntaxModel-SkipID Release 0.29, containing SQL::SyntaxModel::SkipID
0.14, was released on CPAN. This is the last release for this distribution
and module where they both do not explicitly have matching version numbers.
2004-03-22 Darren Duncan <perl@DarrenDuncan.net>
* SQL-SyntaxModel Release 0.28, containing SQL::SyntaxModel 0.15, was
released on CPAN; this is the first release of the SQL-SyntaxModel
distribution.
* SQL-SyntaxModel-ByTree Release 0.28, containing SQL::SyntaxModel::ByTree
0.13, was released on CPAN; this is the first release of the
SQL-SyntaxModel-ByTree distribution.
* SQL-SyntaxModel-SkipID Release 0.28, containing SQL::SyntaxModel::SkipID
0.13, was released on CPAN; this is the first release of the
SQL-SyntaxModel-SkipID distribution.
2004-03-21 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.27, containing SQL::SyntaxModel 0.14 and
SQL::SyntaxModel::ByTree 0.12 and SQL::SyntaxModel::SkipID 0.12, was
released on CPAN. This is the last release of the Rosetta distribution to
contain any version of those 3 modules.
2003-12-21 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.20, containing SQL::SyntaxModel 0.08 and
SQL::SyntaxModel::ByTree 0.08 and SQL::SyntaxModel::SkipID 0.08, was
released on CPAN. This is the first release of any distribution to contain
SQL::SyntaxModel::ByTree, which was split off from SQL::SyntaxModel; this is
the first release of SQL::SyntaxModel following that split.
2003-12-18 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.19, containing SQL::SyntaxModel 0.07 and
SQL::SyntaxModel::SkipID 0.07, was released on CPAN. This is the last
release of SQL::SyntaxModel prior to its split into itself and
SQL::SyntaxModel::ByTree.
2003-09-26 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.18, containing SQL::SyntaxModel 0.06 and
SQL::SyntaxModel::SkipID 0.06, was released on CPAN. This is the first
release of any distribution to contain either SQL::SyntaxModel or
SQL::SyntaxModel::SkipID, which are the results of splitting up the renamed
SQL::ObjectModel.
2003-09-16 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.17, containing SQL::ObjectModel 0.05, was released on
CPAN. This is the last release of any distribution to contain
SQL::ObjectModel; its subsequent release has it renamed and split in two.
2003-06-11 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.13, containing SQL::ObjectModel 0.01, was released on
CPAN. This is the first release of any distribution to contain
SQL::ObjectModel; it is the result of merging 3 pre-existing and renamed
modules, Rosetta::Schema::DataType and Rosetta::Schema::Table and
Rosetta::Schema::View.
2003-06-10 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.12, containing Rosetta::Schema::DataType 0.021 and
Rosetta::Schema::Table 0.011 and Rosetta::Schema::View 0.012, was released
on CPAN. This is the last release of those 3 modules prior to their
three-way merger into SQL::ObjectModel.
2003-03-08 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.06, containing Rosetta::Schema::DataType 0.02 and
Rosetta::Schema::Table 0.01 and Rosetta::Schema::View 0.01, was released on
CPAN. This is the first release of any distribution to contain either
Rosetta::Schema::Table or Rosetta::Schema::View.
2003-03-04 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.05, containing Rosetta::Schema::DataType 0.01, was
released on CPAN. This is the first release of any distribution to contain
Rosetta::Schema::DataType. This is the first release of any distribution
for which some of the existing work that became the SQL-Routine distribution
was in the form of executable code, and not just design documentation.
2003-02-10 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.04 was released on CPAN. This is the last release of
any distribution for which all of the existing work that became the
SQL-Routine distribution was purely in the form of design documentation,
without executable code of any kind.
2003-01-27 Darren Duncan <perl@DarrenDuncan.net>
* Rosetta Release 0.02 was released on CPAN; this is the first release of
the Rosetta distribution. Early versions of the design documentation that
became the SQL-Routine distribution were included.
2003-01-05 Darren Duncan <perl@DarrenDuncan.net>
* DBIx-Portable Release 0.01 was released on CPAN; this is the only release
of the DBIx-Portable distribution. This is the first release of any
distribution to contain design documentation that became the SQL-Routine
distribution.
2002-11-12 Darren Duncan <perl@DarrenDuncan.net>
* Began development on the modules which became the SQL-Routine
distribution as their own entity that is separate from my application, for
open-source distribution on CPAN. The modules were to comprise an
RDBMS-generic DB framework for any application to use.
* Modules based on a template created by h2xs 1.18.
2002-07-28
* Posted the first significant update to the second prototype, which added
an index or cache for data that was expensive to calculate for each page
request, and involved rewriting about a fourth of the perl code.
2002-06-07
* Posted to my website the second public prototype demo of the new
self-proprietary database driven application, whose database schema, user
interface, feature set, and Perl application code was almost completely
rewritten. The new version explicitly had separate modules for database
communication and the web interface logic, with MySQL-specific code and
database generic or data dictionary code in separate files, and separate
files for parts of the web interface. The program used the same generic
CPAN modules as in the first prototype, DBI/MySQL and CGI::Portable.
* This is when I started writing self-contained code components that were
explicitly designed to enable external code that used them to work
seamlessly on multiple database products, and hence 2002 is the start of my
declared copyright date range for SQL::Routine.
2001-11-24
* Posted the last update to the first prototype.
2001-07-12
* Posted to my website the first public prototype demo of a new
self-proprietary database driven application, which is like a cross between
a multimedia metadata catalogue and a repository for historical or
genealogical data. This application contained the first prototypes of code
that ended up in these modules. All of this application's code, for
database communication and web interface logic, was situated in a single
module, with the rest of the program being more generic CPAN modules like
DBI (and DBD for MySQL) and CGI::Portable.
2000-05-17
* Requested MySQL database privileges on my web host so I have something to
start developing, testing and deploying database driven applications on.