ironseed 0.2.0
Breaking Changes
- Avalanche behavior and observed randomness were improved by
including additional bit mixing at the end of hashing. This changes the
calculation of all ironseeds and seed streams, but allows the algorithms
to pass PractRand tests.
- When hashing a string, ironseed now includes the length of the
string in the hash. This avoids input collision for two strings that are
of different length but correspond to the same set of 32-bit values due
to padding.
New Features
- New
with_ironseed()
and local_ironseed()
functions to temporarily change the .Random.seed
state.
- New streaming API:
ironseed_stream()
function returns a
function that can be called multiple times to generate a stream of
output seeds. There are also new with_ironseed_stream()
and
local_ironseed_stream()
functions as well.
ironseed()
gained a new methods
parameter
to control which methods are used for input seeds. The following methods
are supported (in order of default precedence)
- “dots” extracts one or more seeds from arguments passed to
ironseed()
.
- “args” extracts one or more seeds from the command line.
- “env” extracts a seed from an environment variable named
“IRONSEED”.
- “auto” extracts seeds from local sources of system entropy.
- “null” uses a default seed.
- Automatically generated ironseeds will now use also use entropy from
hostnames and cluster job ids (if present).
Miscellaneous fixes and
features
- New
set_ironseed(...)
function to wrap
ironseed(..., set_seed = TRUE)
- New
create_ironseed()
function to construct an ironseed
from objects. This function has a stricter interface than
ironseed()
and is useful for internal usage.
ironseed()
now requires ...
arguments to
be unnammed.
- Better support for MacOS and PowerPC (from Sergey Fedorov @barracuda156)
- Better support for *BSD systems
- Better support for big-endian systems
ironseed 0.1.0