PDB TODOs

More docs. More comments. More tests.

Merge in nested transactions
Support for query profiling

Docs:

Write caching docs
class map (w/ complete doc comments)
how to use pdb query
guide to conditions syntax
installing the compatibility layer
Building your own PdbModel
UUID building
how to not use SQL_CALC_FOUND_ROWS (and why)

Tests:

utilities
Pdb
PdbHelpers
PdbQuery
- SQL building
- executors
PdbSync
PdbCache
Nested transactions

performance tracing

remove PdbModel? (too specific, move it to Bloom I guess)

Adapters:

rename drivers -> adapters
add postgres adapter
sqlite dependent keys

PdbCondition:

make everything use named binds
some sort of magical numeric to number binds thingy

PdbQuery:

select should be keyed by alias
orderBy() and groupBy() should accept strings
add addParams() to (after adding enforced named binds)

Schema stuff:

add 'autoinc-start' per table
add 'secret=1|0' per column
- Records a 'secret' tag in the column comment
- Informs export behaviour
put db_struct.xsd somewhere online
tie down boolean properties to 0|1

Refactor:

rename extended methods (fieldList, indexList)
somehow abstract out driver specific stuff for PdbSync

New stuff:

should return row keys lower-cased? via config?
Pdb::savepoint() for nested transactions
Pdb::transact() should return a transaction ID
- commit() requires this ID as an arg
- forces the user to manage the transaction lifecycle
- rollback prevents any commits
  - throws an error? returns bool?
- can do nested transaction via savepoint()

Fixes:

Pdb::find() should respect 'date_deleted'
sync doesn't remove indexes when renaming a table (needs additional sync)

Document UUIDs somewhere:

'PDB' has a single namespace (uuidv4)
each table then hashes that with 'schema.table.id'
eg. 'twee.users.123'

Maybe:

Make UUID v4 or v5 toggle-able. Maybe.

Uuid 4:

always unique
not deterministic

Uuid 5:

deterministic, hurrah!
Although not reversible (ID -> UUID, UUID !-> ID)
But requires two writes per new record