| Rev |
Age |
Author |
Path |
Log message |
Diff |
| 42208 |
6955 d 3 h |
murdoch |
/branches/djm-source/ |
Recreate branch from trunk |
|
| 41939 |
6984 d 3 h |
ripley |
/trunk/src/main/ |
revert +1 for now |
|
| 41926 |
6985 d 1 h |
ripley |
/trunk/ |
revert R_alloc to char * since C++ seems to care.
remove surplus +1 in R_alloc as it is masking array overruns.
tidy up messages. |
|
| 41916 |
6985 d 11 h |
ripley |
/trunk/src/main/ |
use void* |
|
| 41914 |
6985 d 11 h |
ripley |
/trunk/ |
use void *R_alloc(size_t, int) for 64-bit OSes with 32-bit longs |
|
| 41901 |
6987 d 8 h |
ripley |
/trunk/src/ |
use void* for generic memory pointers
add back-compatibility to R_AllocStringBuffer, with a warning |
|
| 41899 |
6987 d 9 h |
ripley |
/trunk/ |
tidy up and document R_AllocStringBuffer
make use of it, or tidy up usage |
|
| 41894 |
6989 d 2 h |
ripley |
/trunk/ |
deriv was returning double not int dims
make use of mkString and ScalarXXX to simplify code |
|
| 41869 |
6990 d 6 h |
ripley |
/trunk/src/ |
a few more 'const' in declarations |
|
| 41807 |
6993 d 0 h |
rgentlem |
/trunk/ |
patches from Seth F for the CHARSXP hashing changes
Seth says:
const char fixes
CHAR() internal macro now returns (const char *)
Add CHAR_RW: writable CHARSXP access for internal R use only
const char fixes for methods package
Add NEWS item and some doc to R-ext for CHARSXP changes |
|
| 41767 |
6996 d 18 h |
rgentlem |
/trunk/src/ |
more patches from Seth F, for the CHARSXP hashing
Seth says:
This patch series teaches R_StringHash how to handle the encoding
information stored in the general purpose bits of CHARSXPs. A new
interface for CHARSXP creation is added, mkCharEnc that allows for
specifying the encoding.
The patch series also captures more CHARSXPs in the global cache. In
particular, character.c:mkChar2 and gram.y:mkChar2 both use the cache
now.
The string hashing in scan.c has been removed as it no longer is
needed.
The CallocCharBuf macro gets a small fix (extra parens)
And last, but not least:
CHAR() from package code now returns (const char *)
Since CHARSXPs should no longer be modified in place, we can make
CHAR() return (const char *) so that compilers can warn or error about
improper modification. Thanks to Herve Pages for the suggestions. |
|
| 41715 |
7002 d 10 h |
ripley |
/trunk/ |
improvements to error/warning messages |
|
| 41686 |
7005 d 3 h |
ripley |
/trunk/ |
more improvements to warning/error messages |
|
| 41665 |
7006 d 4 h |
ripley |
/trunk/src/main/ |
update count of hash chains in use |
|
| 41636 |
7009 d 1 h |
ripley |
/trunk/src/ |
tidy up inlining some more, including using macros where possible (more widely
usable) |
|
| 41495 |
7019 d 20 h |
rgentlem |
/trunk/src/ |
a large number of changes to alter the behavior of CHARSXPs so that
they are unique and hashed. This will have some effect on run time, but
for any application which uses a lot of names (like big arrays with
row or column names) the savings are often impressive |
|
| 41002 |
7056 d 22 h |
jmc |
/trunk/ |
use DUPLICATE_ATTRIB instead of SET_ATTRIB to copy object, S4 bits correctly |
|
| 40681 |
7108 d 0 h |
ripley |
/trunk/src/main/ |
update |
|
| 40680 |
7108 d 1 h |
ripley |
/trunk/src/ |
modifications for the write barrier test |
|
| 40672 |
7108 d 8 h |
ripley |
/trunk/ |
start marking encoding of CHARSXPs |
|
The R Project SVN
R