std.hash: More questions

[Johannes Pfau]

Code:
https://github.com/D-Programming-Language/phobos/pull/646
Docs:
http://dl.dropbox.com/u/24218791/d/phobos/std_hash_hash.html
http://dl.dropbox.com/u/24218791/d/phobos/std_hash_crc.html
http://dl.dropbox.com/u/24218791/d/phobos/std_hash_md.html
http://dl.dropbox.com/u/24218791/d/phobos/std_hash_sha.html

I just had another look at my initial std.hash design, and I realized
that the API could be simplified a little:

There's a reset function that's implemented in every hash. For sha1,
md5, crc32 it only forwards to the start function though. So I'm not
sure how useful this function is or if it should be dropped.

Advantages of keeping it:
* 'reset' better documents what's done than 'start' if the hash has
  already processed data
* Are there hashes which can implement a reset function in a faster way
  than calling start again?

Cons:
* Adds an additional function which probably isn't necessary


The start function is probably not needed as well. Tango doesn't have a
start function or something similar, but it could use constructors for
this (I only looked at docs, not code). We can't use constructors, so a
start function would be necessary for advanced initialization. But do
we actually need that advanced initialization? SHA1, MD5 and CRC32 just
do a "this = typeof(this).init" so a start function isn't necessary
here.

Advantages of keeping it:
* Are there hash algorithms which need some sort of complex
  initialization which can't be done with .init / default values?
* If we drop both start and reset the only way to reset the internal
  state is calling finish. This might be a little less efficient than a
  start/reset method.

Advantages of dropping it:
* Using hashes is easier, no need to call 'start' before hashing data

I think someone more familiar with hash functions than me needs to
answer the "do we need start/reset functions" questions.


API question:

CRC32 sums are usually presented as a uint, not a ubyte[4]. To fit the
rest of the API ubyte[4] is used. Now there's a small annoying detail:
The CRC32 should be printed in LSB-first order.
When printing an uint like this, that works well:
writefln("%#x", 4157704578); //0xf7d18982
but this doesn't:
toHexString(*cast(ubyte[4]*)&4157704578); //8289D1F7

I can't change toHexString as it's used for all hashes and it's correct
for SHA1, MD5, ...
So I currently use bswap in the CRC32 finish() implementation to fix
this issue.

Now the question is should I provide an additional finishUint function
which avoids the bswap? 


Implementation issue:

The current implementation of SHA1 and MD5 uses memcpy which doesn't
work in CTFE IIRC and which also prevents the code from being pure.
I could replace those memcpy calls with array copying but I'm not
sure if memcpy was used for performance, so I'd like to keep it as long
as we have no performance tests.