currency wire format

[jargoman]

BinarySerializer, write currency.

Your code is correct but ripple rejects "floating point" values whose mantissa doesn't have 16 digits.

Lets say I want to send the value 1.23.
Instead of the way you would expect.
mantissa = 123; // unscaled value
scale = 2; // or negative 2 in some implementations.

ripple expects it to be in wire format
mantissa = 1230000000000000;
scale = 15; // again may be -15

Both values are equal but only the second is accepted by rippled

1230000000000000 * 10 ^ -15 == 123 * 10 ^ -3; // statement is true

This is why the max and min scale have a difference of 16
80 - 96 = 16.
and is also why 54 bits is allocated for the mantissa
9999999999999999 // 54 bit number.

It took me three days to figure out why your code was correct but returning "invalid currency value"

[sublimator]

A good way of testing implementations of binary parsing/reading is to implement a ShaMap and download some ledgers, calculating the transaction_hash. You can retrieve the transactions in binary form from rippled.

[pmarches]

Ah! That would explain why the C++ code iterates *10 and /10 to make the
mantissa 16 digits long. I tried to be smart in avoiding that loop. Do you
have a testcase or transaction that reproduces the problem?

sublimator: If we read the TX from the binary fromat, it actually computes
the hash correctly, since the BigDecimal will be created with a 16 digit
mantissa. The problem lies when we create a new BigDecimal from scratch
that does not have the 16 digit mantissa.

Thank you to both of you for this,

Philippe

On Wed, Feb 26, 2014 at 1:45 PM, sublimator notifications@github.comwrote:

A good way of testing implementations of binary parsing/reading is to
implement a ShaMap and download some ledgers, calculating the
transaction_hash. You can retrieve the transactions in binary form from
rippled.

Reply to this email directly or view it on GitHubhttps://github.com//issues/20#issuecomment-36181238
.

[sublimator]

You can load from the rippled json too (for sanity testing)

[sublimator]

Actually a guy on the ripple forum posted a heap of full ledgers