1 |
cvsjoko |
1.1 |
Abstracted Types (Stig) |
2 |
|
|
----------------------- |
3 |
|
|
|
4 |
|
|
DB needs a set of types representing the most commonly used types in |
5 |
|
|
all backends. This type set could also be geared towards integration |
6 |
|
|
with things like XML-RPC/SOAP implementations, HTML form classes, etc. |
7 |
|
|
|
8 |
|
|
Real Query Parser (Stig) |
9 |
|
|
------------------------ |
10 |
|
|
|
11 |
|
|
With a real query parser, DB can implement more of its portability |
12 |
|
|
based on the query, instead of having support functions for |
13 |
|
|
everything. One example would be LIMIT, another "INSERT |
14 |
|
|
... RETURNING". |
15 |
|
|
|
16 |
|
|
Portable transactions (Stig) |
17 |
|
|
---------------------------- |
18 |
|
|
|
19 |
|
|
If DB can parse queries enough to determine what tables are affected |
20 |
|
|
by queries, it should be possible to make a replayable transaction |
21 |
|
|
log. GNOME uses an XML format for configuration data that lets you |
22 |
|
|
checkpoint state once in a while, and revert to that state later. |
23 |
|
|
With a similar approach for transactions in DB we can implement |
24 |
|
|
portable transactions and checkpointing even for the databases that |
25 |
|
|
don't support them. |
26 |
|
|
|
27 |
|
|
|
28 |
|
|
Error reporting clean-up/debug (Tomas) |
29 |
|
|
------------------------------------- |
30 |
|
|
Now each driver has its own raiseError method, common has a raiseError and |
31 |
|
|
DB has a DB_error class and its own isError() method. This error stuff |
32 |
|
|
overhead could be simplified with only one raiseError, droping the DB Error |
33 |
|
|
class and also the DB::isError() (use the PEAR.php ones instead). |
34 |
|
|
Other idea could be to add a system for allowing people access to all the |
35 |
|
|
queries sended by PEAR DB to the backend. Also a new PEAR_ERROR_DEBUG |
36 |
|
|
flag that automatically (show|triggers) debug info, perhaps |
37 |
|
|
with a new PEAR_(Warning|Debug) object. |
38 |
|
|
|
39 |
|
|
Quote clean-up (Stig) |
40 |
|
|
--------------------- |
41 |
|
|
1. Keep quote and quoteString, but move quoting of strings back into |
42 |
|
|
quoteString and make quote call it for strings. |
43 |
|
|
|
44 |
|
|
2. Add an optional "operator" parameter to quote that is one of "=", |
45 |
|
|
"<", ">" or "<>" that will be inserted in front of the quoted value |
46 |
|
|
unless it is NULL, in which case it will be converted to "IS" (for |
47 |
|
|
"=") or "IS NOT" (for the others). |
48 |
|
|
|
49 |
|
|
Auto free statements (Tomas) |
50 |
|
|
---------------------------- |
51 |
|
|
By setting a param in query() or for the hole DB instance, PEAR DB |
52 |
|
|
could auto-free results in DB_result->fetch(Into|Row) when the driver |
53 |
|
|
returns false. |
54 |
|
|
|
55 |
|
|
Datatypes in prepare syntax (Tomas) |
56 |
|
|
----------------------------------- |
57 |
|
|
Extend the actual prepare/execute placeholders to support data types, both |
58 |
|
|
to check the data introduced to the query and to "cast" the result |
59 |
|
|
to native php data types. Ex: |
60 |
|
|
|
61 |
|
|
$sql = "INSERT INTO table VALUES ({{int(4)}}, {{bool}}, {{date('Y-m-d')}})"; |
62 |
|
|
$row = $db->query($sql, array(8, 't', '2001-04-1')); |
63 |
|
|
|
64 |
|
|
Format: {{<data_type>(<param1>,<param2>)}} |
65 |
|
|
|
66 |
|
|
"param" could be the max lenght of the data, date formats, not_null |
67 |
|
|
checks or default values. |
68 |
|
|
|
69 |
|
|
Other ideas could be: |
70 |
|
|
|
71 |
|
|
1) |
72 |
|
|
$sql = "INSERT INTO table VALUES (?, ?, ?)"; |
73 |
|
|
$sth = $db->prepare($sql, array('int(4)', 'bool', 'date'); |
74 |
|
|
$res = $db->execute($sth, array($a, $b, $c); |
75 |
|
|
|
76 |
|
|
2) |
77 |
|
|
$sql = "INSERT INTO table VALUES (?, ?, ?)"; |
78 |
|
|
$params = array( |
79 |
|
|
0 => array($a, 'int(4)'), |
80 |
|
|
1 => array($b, 'bool') |
81 |
|
|
); |
82 |
|
|
$res = $db->query($sql, $params); |
83 |
|
|
|
84 |
|
|
Auto connect feature (Tomas) |
85 |
|
|
---------------------------- |
86 |
|
|
Add the ability to create for example a light and dump DB object which |
87 |
|
|
will only set up the connection when needed. With that people could |
88 |
|
|
create the DB object in a common prepend or default file without the |
89 |
|
|
need to waste system resources if the use of the database is finally |
90 |
|
|
not needed. |