KDb/Archives/libCalligraDB

The Task: Move core parts of KexiDB lib from kexi/kexidb to calligra/libs/db (libcalligradb) for reuse by Words' Bibliographic Database module and possibly other places in Calligra.

Status: in master (2.6) since 2012-07-11

Detailed changes:

GUI- or Kexi-related bits such as validators are still in Kexi
Only the SQLite3 driver is moved to libcalligradb, others stay within Kexi
- thus, calligra libs depend on the driver, i.e. depend on SQLite3 (this was the case already for Words and Kexi)
- calligra libs depend on ICU as well for proper unicode handling (which is dependency of Qt 5 anyway)
~~Predicate~~KDb library, as a Qt-only component, will replace functionality of the calligradb lib
sqlitedump tool used for database for interactive database compacting stays in Kexi; if needed could be moved to libs/ too
a few functions from libkexiutils moved to libcalligradb, so libkexiutils is not dependency
for simplicity/API stability, namespace stays as KexiDB
README.PACKAGERS file updated

Plans:

port the biblio db storage from plain QtSQL to libcaligradb (smitpatel)
- with support from jstaniek
- target version: 2.6
port the biblio db storage to Predicate (by who?)
- with support from jstaniek
- target version: 3.0

Minutes from IRC meeting

2012-07-03
requested by jstaniek
attended by boemann, jstaniek, smitpatel

idea: storage for bibliography could re-use libKexiDB instead of libQtSQL
pros:
- simplify the code
- increase code re-use - KexiDB is developed 8+ years already with reuse in mind
- make the database ($HOME/.calligra/biblio.sqlite) accesible to Kexi, so biblio db can be added to the 'Kexi Scientific' family of templates for data re-use and manipulation
- KexiDB has database and table creation API, hides SQL complexity, QtSQL does not (what may result in code complexity in Words)
cons and questions:
- size of libKexiDB as dependency (Biblio already uses SQLite3, which is the biggest dependency compared to either QtSQL or KexiDB anyway)
- we're moving to Predicate in Kexi (but that would happen no earlier than in 3.0, realistically in 3.1 or 3.2), so KexiDB is here for the transition period
- are the APIs stable? (yes, most changes happen in Predicate, moreover KexiDB is Calligra's internal API, so will be updated with the apps and the Biblio code)
- Words cannot depend on Kexi (it wouldn't, relevant parts of KexiDB should be moved to calligra/libs/)

extra points, from Kexi's perspective
- According to plans, Kexi will have biblio-feature db in 3.x anyway as a template, which ideally would depend on Words too. With kexidb reuse we would avoid incompatible two databases, avoid extra work, and avoid two Biblio products.
- For now Kexi lacks templates that are real world and integrated with other Calligra apps. The Biblio initiative helps here.

current state: smitpatel delivered QtSQL-based code for Biblio (within GSoC 2012), jstaniek thinks it has bits (db and table creation) already handled in KexiDB

final agreement has been reached: KexiDB adoption in Biblio
- smitpatel already studied kexi/tests/newapi sample code and claims it looks pretty straight forward for use in Biblio
- Biblio can be still kept small enough after using with KexiDB, in fact codebase size of Biblio will decrease
- Kexi is not exposed to Words' users in any way; conversely - users may find the connection in Kexi if they decide
- jstaniek offers to move relevant part of KexiDB to calligra/libs/ ASAP
  - the library renames to calligradb and the dir is calligra/libs/db/ (to avoid references to Kexi)
  - missing bits of the API needed by Biblio can be added (since there's no BC/SC requirement in KexiDB)
- jstaniek offers support for smitpatel in KexiDB use
- jstaniek offers support for porting to Predicate when it arrives
- jstaniek offers maintaining the libcalligradb for Qt 5.0
- jstaniek proposes to keep KexiDB namespace in libcalligradb however since it defines API which is heavily used in Kexi (3200+ times); after moving to Predicate it won't be a problem
- all devs want to avoid changes of the db format so the current QtSQL-based code even if released, will be marked as experimental, before moving to KexiDB