Revision history for Perl extension Embedix::DB.

+ means it still needs to be done
- means it's done

beyond 1.00

+ move node.help and node.prompt to another table (node_description) in order to facilitate localization. + go back to California

1.00 TARGET

+ be able to configure an Embedix distribution (another module) + add arch table?

      If I do, the architectures should be exactly the same as the ones
      found in /usr/src/linux/arch/*.

+ figure out a way to do dependency checking (another module) + figure out a way to do conflict resolution (another module) + figure out how I want to deal with requires, requiresexpr, && if + find a clever use for AUTOLOAD
+ should root nodes have an entry in node_parent w/ parent_id = 0? + cloneDistro(arch => $whatever);
+ handle attributes I am currently ignoring:

      choicelist
      trideps 
      license
      if - the only non-trivial one
    + implement unrelateNode() and maybe think of a better name
      while I'm at it.
    + eliminate unnecessary prepares and use $dbh->do() when
      appropriate.

+ write a test for getComponentList() + I need to remove workOnDistro() from the public API and make it part of the constructor instead. + implement deleteNode() and a test for it. + implement deleteNodeDependencies() and a test for it.

      This will just be a front-end to the 
      plpgsql function node_dependencies_delete();
    + using advanced perl OO-magick, make a subclass of Embedix::ECD
      that has delayed node retrieval...  or maybe I should add hooks
      to Embedix::ECD that makes it friendly towards delayed loaders.
      That would be PIMP!

0.05

• implemented getDistroList() which has similar behaviour to getComponentList();
• document getDistroList
• added stub for admin functionality to bin/ebx

0.04 Sun Feb 25 09:51:04 MST 2001

• [Pg] factor a function out of node_dependencies_delete() such that there will exist a function called node_aggregates_delete(id) that deletes aggregates but not child nodes;
• removed DB/CML2.pm, because it's going to go with the dep-handling module (whatever I decide to call it).
• in the methods that insert aggregate attributes, I have it skip blank lines -- it used to insert them.
• in the methods that insert aggregate attributes, it is now able to prevent duplicates from being inserted. Currently, postgres itself prevents an insertion, but in doing so, it totally invalidates the current transaction. This change pre-empts that behaviour.
• s/insertBuild_vars/insertBuildVars/g;
• finished implementing getNodePath() w/ a path_cache even.
• getComponentList() should be OK, too
• changed t/00_pg.t to clean up more rigorously
• removed getHierarchy
• s/arch/board/g; Specifying the board is more specific, and it flows better with the direction Stuart is taking with building. This affected DB/Pg.pm, sql/pg_schema.sql, and t/00_pg.t -- "make test" survived.
• hashrefFromECD() no longer has a distro_id, because we're going to be using the node_distro table. distro_id was also removed from the node table.
• selectNode() had its references to distro_id removed "make test TEST_VERBOSE=1" looked ok.
• added relateNode() to make insertions into the node_distro table.
• I need to make insertions into node_distro in: -addDistro -cloneDistro # of course -insertNode
• I need to join node_distro in -getNodePath -selectNode -getComponentList
• SERIOUS ISSUE: If I allow nodes to be shared between distros, then I will need to introduce a node_distro table similar to the node_parent table that defines a mapping between node and distro. Although a node may be shared between many distros, it doesn't necessarily have to be shared. This also means getting rid of the distro_id column in the node table. cloneDistro() is impossible to implement until this is done. ...but I did it.
• fixed scripts in bin/ to use 'board' instead of 'arch'.
• fixed typo in select statement for getComponentList.
• [Pg] "conflicts" attribute added to node table. I hope it's not one of those stealth multi-line attributes.
• changed getNodePath() to prepend '/' on all paths.
• implemented unrelateNode()
• massive documentation updates for Embedix::DB
• changed the *size attributes into int arrays. I'm able to evaluate the mathematical expressions before inserting, too.
• [Pg] added "if" column to node table
• made t/01pg.t fail gracefully.

0.03 Tue Jan 16 04:22:43 MST 2001

• renamed addECD to updateDistro(ecd => $ecd, parent_id => $id). parent_id is optional.
• made Embedix::DB::Pg::hashrefFromECD be able to handle range attributes. I implement them as PostgreSQL arrays in the schema.
• [Pg] created indexes (indices?) on node for node_class and node_name
• [Pg] created an index on node_parent for parent_id
• [Pg] added an arch column for distro
• [Pg] added a license and node_license table
• [Pg] removed all the user_* tables for now, because I have to think about them more.
• [Pg] addDistro and/or workOnDistro may have a subtle bug I wrote some more test cases, and everything seemed to be OK, though. It even behaves right in edb_populate.pl where I originally had the problem. I think the original errors may have been caused by not cleaning up some tables properly. workOnDistro() was probably fine all along.
• use Carp;
• be able to add aggregate attributes qw(provides build_vars keeplist)
• can add provides and requires aggregate attributes
• make addDistro and workOnDistro care about arch
• implemented getHierarchy
• preserve the correct $dbh->errstr in those parts that say croak($dbh->errstr . $q); Added the method, rollbackAndCroak, to distill this idiom and save space.
• [Pg] created a function and trigger combo to recursively delete node dependencies when a node is deleted.
• [Pg] added tables for aggregates: choicelist build_vars
• [Pg] be able to insert requires attribute into node table
• [Pg] in insert(Provides|Keeplist|Build_vars), carp instead of croak when a primary key is already taken.

0.02 Mon Jan 1 18:55:20 MST 2001

• finally wrote a schema I could play around with.
• decided that Embedix::DB will be a factory class that instantiates instances of Embedix::DB::* backend objects. These backend objects will NOT be subclasses of Embedix::DB. Embedix::ECD and friends may get reorganized similarly.
• added methods: qw(new workOnDistro addDistro addECD) to Embedix::DB::Pg
• stole build_update_statement an build_insert_statement from DBIx::Hotel
• added Embedix::DB::Pg::currval()
• added Embedix::DB::Pg::nextval()
• added sql/pg_reset.sql for making it easy to start over
• changed buildInsertStatement to use $dbh->quote
• changed buildUpdateStatement to use $dbh->quote
• added selectNode, updateNode, and insertNode. I went to sleep before testing it.
• I am so fscking good. addECD worked on the first try. damn, I'm good. (wait a few hours for my ego to deflate) There were a few minor bugs, but all the inserts happened w/o dying.

0.01 Sat Dec 9 12:58:40 MST 2000

• original version;