Debian Codegraph @

TO_DO

raw
1 use strict;
2
3 =head1 NAME
4
5 DBD::ODBC::TO_DO - Things to do in DBD::ODBC
6
7 As of $LastChangedDate: 2010-10-08 17:00:31 +0100 (Fri, 08 Oct 2010)$
8
9 $Revision: 10667 $
10
11 =cut
12
13 =head1 Todo
14
15 Better/more tests on multiple statement handles which ensure the
16 correct number of rows
17
18 Better/more tests on all queries which ensure the correct number of
19 rows and data
20
21 Better tests on SQLExecDirect/do
22
23 Keep checking Oracle's ODBC drivers for Windows to fix the Date
24 binding problem
25
26 There is a Columns private ODBC method which is not documented.
27
28 Add support for sending lobs in chunks instead of all in one
29 go. Although DBD::ODBC uses SQLParamData and SQLPutData internally
30 they are not exposed so anyone binding a lob has to have all of it
31 available before it can be bound.
32
33 Why does level 15 tracing of any DBD::ODBC script show alot of these:
34 !!DBD::ODBC unsupported attribute passed (PrintError)
35 !!DBD::ODBC unsupported attribute passed (Username)
36 !!DBD::ODBC unsupported attribute passed (dbi_connect_closure)
37 !!DBD::ODBC unsupported attribute passed (LongReadLen)
38
39 Add a perlcritic test - see DBD::Pg
40
41 Anywhere we are storing a value in an SV that we didn't create
42 (and thus might have magic) should probably set magic.
43
44 Add a test for ChopBlanks and unicode data
45
46 Add some private SQLGetInfo values for whether SQL_ROWSET_SIZE hack
47 works etc. How can you tell a driver supports MARS_CONNECTION.
48
49 Might be able to detect MARS capable with SS_COPT_MARS_ENABLED
50
51 Bump requirement to Test::Simple 0.96 so we can use subtest which
52 is really cool and reorganise tests to use it. 0.96, because it seems
53 to be the first really stable version of subtest.
54
55 Add more Oracle-specific tests - like calling functions/procedures
56 and in/out params.
57
58 Download rpm package from here ->
59 http://download.opensuse.org/repositories/devel:/languages:/perl/openSUSE_11.4/src/
60 and see what changes they are making (especially Makefile.PL) to see if we
61 might need to include them.
62
63 DRIVER and DSN in ODBC connection attributes should be case insensitive
64 and they are not - they are strcmped against DRIVER/DSN - check!
65
66 Some people still have problems with iODBC being picked up before
67 unixODBC. Various ideas from irc:
68
69 <Caelum> mje: what does undefined symbol: SQLGetFunctions mean?
70 <Caelum> mje: do I have iodbc conflicting again?
71 <@Caelum> yeah I think that's it
72 <mje> perl Makefile.PL -x
73 * gfx is now known as gfx_away
74 <Caelum> mje: kinda sucks I have to do that every time, every auto-upgrade borks it
75 <mje> can't you just uninstall iODBC
76 <mje> is this OSX?
77 <@Caelum> it's Debian, I think I have iODBC because amarok needs it
78 <mje> oh, so use -x
79 <ribasushi> mje: what if you just compile a quick program in the Makefile itself, and determine whether you need -x or not?
80 <@ribasushi> I've seen other things to that - e.g. Time::HiRes
81 <mje> could do, I guess I could revisit this - problem was Jens (I think) sent me a patch to prefer iODBC over unixODBC and told me it was required for something - can't remember right now
82 <mje> otherwise, I'd just change the default to unixODBC then iODBC
83 <@Caelum> maybe an env var in addition to -x? then I could just put it in my .zshrc like PERL_DBD_ODBC_PREFER_UNIXODBC
84 <mje> yeah, could do that too
85 <ribasushi> mje: the point is you can't really "prefer" one or the other - you need to check for a specific -dev existence, instead of assuming it's there if the package itself is there
86 <mje> but iODBC and unixODBC dev packages both provide sql.h etc
87 <mje> you should see the code I already have to try and fathom this out - it is massive
88 <ribasushi> mje: by check I meant - compile something small and quick to run
89 <mje> some unixODBC's don't include odbc_config, some do, some do but are too old and don't have switch fred
90 <mje> to compile correctly I need to find odbc_config or iodbc_config so catch 22
91 <@ribasushi> but here's the kicker - you compile "something" - if it works - you found things correctly
92 <@ribasushi> if the compile/execution fails - you "found" wrong
93 <mje> I don't think it is that simple - I need to know it is iODBC or not and a whole load of other things
94 <mje> there is a lot of history in the tests for supporting older systems
95 <@ribasushi> https://metacpan.org/source/ZEFRAM/Time-HiRes-1.9724/Makefile.PL#L334
96 <ribasushi> mje: ^^ I was referring to stuff like this
97 <@ribasushi> without changing the heuristics at all, simply looping the whole thing with different params depending on the config outcome
98 <@ribasushi> it's ugly, but can end up rather effective
99 <mje> I'll look at it again - added to TO_DO - where does try_compile_and_link come from?
100 <ribasushi> mje: it's all part of this makefile, it's a good read rather clean
101 <mje> ignore - sorry
102 Jens suggested Config::AutoConf
103
104 Cancel is not documented
105
106 ODBC 3.8:
107 http://msdn.microsoft.com/en-us/library/ee388581%28v=vs.85%29.aspx