incomplete_types revision 362181 
1The question has been asked multiple times, "Why is APR using Incomplete
2types?"  This document will try to explain that.
3
4Incomplete types are used in APR because they can enforce portability, and
5they make the APR developers job easier, as well as allowing APR to use native
6types on all platforms.  Imagine a scenario where APR wasn't using incomplete
7types.  The ap_file_t type would have to be defined as:
8
9typedef struct ap_file_t {
10    ap_pool_t *pool
11    char *fname;
12    int eof_hit;
13    int pipe;
14    ap_interval_time_t timeout;
15#ifdef WIN32
16    HANDLE file_handle;
17    DWORD dwFileAttributes;
18#elif defined(OS2)
19    HFILE filedes;
20    HEV PipeSem
21#else
22    int filedes;
23    int ungetchar;
24#endif
25
26#ifndef WIN32
27    int buffered;
28    ap_int32_flags
29    int isopen;
30   
31    /* Stuff for buffered mode */
32    char *buffer;
33    int bufpos;
34    unsigned long dataRead;
35    int direction;
36    unsigned long filePtr;
37    ap_lock_t *mutex; 
38#endif
39} ap_file_t;
40
41This captures the essence of what is currently being defined for ap_file_t
42using incomplete types.  However, using this structure leads developers to
43believe that they are safe accessing any of the fields in this structure.
44This is not true.  On some platforms, such as Windows, about half of the
45structure disappears.  We could combine some of these definitions with
46macros, for example:
47
48#ifdef WIN32
49#define filetype HANDLE
50#elif OS2
51#define filetype HFILE
52#else
53#define filetype int
54#endif
55
56And then in the definition for ap_file_t, we could say:
57    filetype filedes;
58
59This gets rid of some of the complexity, by moving it off to the side, but
60it is still not safe for a programmers to access the filedes field directly 
61outside of APR, because the programmer has no way of knowing what the actual 
62type is.  So for example printing the filedes using printf would yield wildly 
63varying results on Windows and OS2 when compared to Unix.
64
65Another option also presents itself.  Stick strictly to POSIX.  This means
66that all code can be shared on any POSIX compliant platform.  The problem
67with this is performance.  One of the benefits to APR, is that it allows
68developers to easily use native types on all platforms with the same code.
69This has proven to provide a substantial performance boost on most non-Unix
70platforms.
71
72Having said all of that, sometimes incomplete types just don't make sense.
73For example, the first implementation of time functions used incomplete types,
74which added a layer of complexity that turned out to be unnecessary.  If
75a platform cannot provide a simple number that represents the number of seconds
76elapsed since a specified date and time, then APR doesn't really want to
77provide support for that platform.
78
79APR is trying hard to provide a balance of incomplete and complete types, 
80but like all things, sometimes the developers make mistakes.  If you are
81using APR and find that there is an incomplete type that doesn't need to be
82an incomplete type, please let us know, we are more than willing to listen
83and design parts of APR that do not use incomplete types.
84
85