| 1 |
- stack and list classes should have operator=() !!! |
| 2 |
* This will be rather messy to implement... as it will require operator=() on |
| 3 |
all the nodes, which in turn may require operator=() on the contained |
| 4 |
objects. This is OK for the e-containers; but the o-containers will then |
| 5 |
have to force contained objects to implement operator=(). |
| 6 |
- list (perhaps also stack) classes need a cut() operator that allows more |
| 7 |
efficient destructive list copying/attaching. How about something similar to |
| 8 |
Perl's splice? |
| 9 |
- revise existing libs to use it (our libs REALLY need non-b0rked exception |
| 10 |
handling; error.h is only good for C) |
| 11 |
- implement error-handling in stack templates!! |
| 12 |
* must use new exception.* classes. |
| 13 |
- implement target to build tar.gz archive, with semi-automatic MANIFEST |
| 14 |
generation, etc.. |
| 15 |
+ implement standardized library error-code passing. This will probably be |
| 16 |
based on the error_d/error_t strategy using error codes that are unified |
| 17 |
with their respective error messages. |
| 18 |
+ re-design the interface for the FILE hierarchy. You'll also need to consider |
| 19 |
how to maximize backward compatibility while fixing flaws in the original |
| 20 |
design. |
| 21 |
+ implement iterators for stack templates! |
| 22 |
+ look into why *iter++ doesn't compile, in list.h. Possibly due to |
| 23 |
operator++ return type: maybe we shouldn't specify return type?? |
| 24 |
** No, the problem is, operator++() must be overloaded to return the |
| 25 |
derived class object rather than the base class object. This is NOT |
| 26 |
automatically done! |
| 27 |
+ implement doubly-linked lists! |
| 28 |
X port SPLITSRC. Or perhaps re-implement it, since the existing code is too |
| 29 |
messy. |
| 30 |
* Forget it. It's not that useful anymore... or at least, if you want it, |
| 31 |
you should redesign with a less b0rked design! |
| 32 |
X design a set of basic print-formatting operations, such as typesetting |
| 33 |
integers, etc., that respects buffer bounds. |
| 34 |
* Finally the libc people have caught on, and have implemented niceties like |
| 35 |
snprintf() and vsnprintf(). |
| 36 |
+ re-organize this messy directory! |
| 37 |
+ convert to use Cons |
| 38 |
+ separate C and C++ libs (too messy otherwise). |
| 39 |
X file.h probably needs re-designing... esp. the read/write interface |
| 40 |
(though fread/fwrite seems cumbersome, they have better error behaviour |
| 41 |
when it comes to file size). |
| 42 |
* the file.* hierarchy has been deprecated. Design is obsolete. |
| 43 |
X re-design config file module! |
| 44 |
* will need redesign from scratch; old design is obsolete. |
| 45 |
X write README file for this dir to describe lib usage. In particular, |
| 46 |
note that alternative implementations of a module can be selected by |
| 47 |
specifying the implementation dir in the #include path, eg.: |
| 48 |
#include <ansic/pfile.h> |
| 49 |
#include <linux/config.h> |
| 50 |
This mechanism has many flaws, though... we should come up with a better |
| 51 |
way of managing selection of alternative implementations! |
| 52 |
* Split up into C/C++ libraries. We'll worry about alternative |
| 53 |
implementations in the future. At any rate, it should be much easier |
| 54 |
under Cons. |
| 55 |
X adapt DBUF module code to work with new error-code passing scheme. |
| 56 |
* DBUF is obsolete. Probably should adapt idea into a better design. |
| 57 |
X clean up code for buffered file classes (BFILE module). You need to make it |
| 58 |
more secure! |
| 59 |
* the FILE hierarchy is obsolete. |
| 60 |
+ incorporate better exception-handling class from atom4/x |
| 61 |
+ probably need to clean it up a bit before we put it in here |
| 62 |
+ at least need some form of locus field |
| 63 |
+ *could* just implement this as class hierarchy and just have the locus |
| 64 |
as a convention rather than an enforced rule. |