"base64-encoded little-endian doubles" in java-CodePudding

I am currently trying to use java to create requests for the DWave solvers. At the moment I am trying to wrap my head around the JSON-encoding part. Especially how the parameters of the problem are to be transmitted.

The description just reads: base64-encoded little-endian doubles (see https://docs.dwavesys.com/docs/latest/c_rest_api_5.html) which is only little info. To try and get info I tried to decode the String in the example. (The one starting with AAAAAAAA4L8AAAAAAADwPwAAAAAAAAAAAAAAAAAA H amZmZmZnJP....). Trying to decode it made only a IllegalArgumentException, or when shortening the String it sometimes returned some gibberish.

Does anyone have an idea how this works?

(I tried to look at the ocean-sdk source code however I'm not good enough at python to read it)

CodePudding user response：

The example in the documentation contains multiple base64-encoded lines:

AAAAAAAA4L8AAAAAAADwPwAAAAAAAAAAAAAAAAAA H amZmZmZnJPwAAAAAAAPh/AAAAAAAA H8A
AAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAA
AAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAA
AAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAA
AAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAA
APh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA
 H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4
fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/
AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8A
AAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAA
AAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAA
AAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAA
AAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAA
APh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA
 H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4
fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/
AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8A
AAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fw==

The first line decodes to the following:

$ printf 'AAAAAAAA4L8AAAAAAADwPwAAAAAAAAAAAAAAAAAA H amZmZmZnJPwAAAAAAAPh/AAAAAAAA H8A' | base64 -d | xxd
00000000: 0000 0000 0000 e0bf 0000 0000 0000 f03f  ...............?
00000010: 0000 0000 0000 0000 0000 0000 0000 f87f  ................
00000020: 9a99 9999 9999 c93f 0000 0000 0000 f87f  .......?........
00000030: 0000 0000 0000 f87f 00                   .........

A double comprises 8 octets, so the first double value is 0000 0000 0000 e0bf. Interpreting this as an IEEE 754 floating-point double gives you: 2.8426e-319.

The line breaks have to be removed to give you the full base64-encoded input. I don't think the line breaks are required when serializing, but you can simply add them after you have encoded as base64 (if they are missing).

To encode an array of doubles, you have to first convert to a byte array which can then be fed to the Base64 encoder.

Parsing the full base64 string will give you 128 double values:

$ printf 'AAAAAAAA4L8AAAAAAADwPwAAAAAAAAAAAAAAAAAA H amZmZmZnJPwAAAAAAAPh/AAAAAAAA H8A
AAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAA
....
AAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fwAAAAAAAPh/AAAAAAAA H8AAAAAAAD4fw==' | base64 -d | xxd -p -c0 | fold -w16
000000000000e0bf
000000000000f03f
0000000000000000
000000000000f87f
9a9999999999c93f
000000000000f87f
000000000000f87f
# ... value repeated ...
000000000000f87f