Hello, Francisco Reyes wrote: > > I am seeking feedback/comments/recommendations on an abstraction layer to > use for our development. So far I have found that the following layers > exists:PHPLib, ADODB, PEAR DB, MetatData > > I have only seen examples from Pear and will try to see examples from the > others, but I am wondering if anyone who has tried them could provide > feedback. If you are looking for a database abstraction package that provides true database independence to your applications or components, you may want to try Metabase (not Metadata). Metabase provides true abstraction of database types. This means in practice that from the point of view of the applications it always handles data formatted in the same way regardless of how it is represented in the database. For instance, if your application deals with dates, it only sees them formatted according to the ISO 9601 standard (YYYY-MM-DD) even if the underlying databases represents dates in other format. Metabase will take care of datatype conversions if necessary when storing or retrieving data from the database. A direct consequence of this, is that if you switch between databases you will not need to touch your application. Metabase was thought not only for database independence but also for Web database programming in particular, so it offers some features that take in consideration the constraints of Web programming. Metabase pioneered the abstraction of things like providing a way to offer the LIMIT clause functionality but in a way that works in all supported database besides MySQL ( PostgreSQL, Oracle, MS SQL server, Interbase, Informix, MiniSQL, ODBC, etc..). This is important for display query results split in multiple pages. This class is based on Metabase and takes advantage of this feature to display query results in HTML tables eventually with a navigation bar that lets you go back and forth between the different pages of the results. http://phpclasses.upperdesign.com/browse.html/package/130 Metabase also pioneered the use of sequences in such way that it works with databases that do not support sequences natively but support auto-incremented table fields. This is important to insert entries with unique id fields that are usually the table primary keys. The sequence generation is always an atomic process, so no matter how many entries may be entered into a table by simultaneous users, it is always done in a consistent manner. Another key feature of Metabase is the ability to install and maintain database schemas. All you need to do is to describe your database schemas in a custom XML file that specify the names and properties of your tables, fields, indexes and sequences, all in a database independent format. Metabase manager class is able to process that schema description and install it for you, no sweat. If some day you need to change that schema, and often you need to, you just need to change the schema description file. Then Metabase manager is able to process it and compare with the previously installed schema to figure the changes that were done and applying them safely to your schema. This is done without affecting the data that was added after the first time the schema was installed or updated for the last time. A new release of Metabase was made just yesterday. It provides some features that users were requesting like the ability to customize error handling. It also provides yet another innovating features, which is the SQL REPLACE implementation. This is a statement like SQL INSERT except that if the values of the primary keys match the ones of an existing row, it will update that row instead of inserting a new one. AFAIK, only MySQL provides SQL REPLACE command, but Metabase provides a suitable emulation that works the same way using transactions so it works also with PostgreSQL, Oracle, MS-SQL server, Informix, Interbase, etc.. One important new feature is a set of new functions to fetch result set data in a single call. These functions were contributed Lukas Smith and Christopher Linn and are meant to end the complaints of some users that claim that other PHP database abstraction packages are faster than Metabase because they can fetch whole rows of data at once. These new functions in Metabase should end those complaints. Anyway, it is better for you to try it yourself. Metabase documentation is through and very complete. There is also a small tutorial to get you started quickly. If for some reason Metabase does not offer some facility you need, you may want to contact me to tell about it, or better implement it yourself and contribute your work to features yourself in the hall of fame of Metabase contributors. Metabase is free and is distributed with a BSD Open Source licence. For those that don't know Metabase, you may find it here: http://phpclasses.UpperDesign.com/browse.html/package/20 Regards, Manuel Lemos
