Shorter encoding for numbers (breaking change) #9
Conversation
|
I chose '_' as the end character, since it's higher than numbers, and makes the output more readable. |
|
Yes indeed, numbers where encoded as "fixed length record". This was needed for the digit flipping trick to work on negative numbers. Adding a char superior to 9 to the mantissa of a negative number should work, that's a nice idea ! |
codec/number.js
Outdated
| @@ -14,8 +16,8 @@ exports.encode = function (number) { | |||
|
|
|||
| var splitScientificNotation = number.toExponential().split('e'); | |||
| var exponent = Number(splitScientificNotation[1]) + 500; | |||
| var mantissa = splitScientificNotation[0] + (splitScientificNotation[0].indexOf('.') === -1 ? '.' : '') + '0'.repeat(20); | |||
| var encoded = 'E' + padStart(String(exponent), 3) + 'M' + String(mantissa); | |||
| var mantissa = splitScientificNotation[0] + (splitScientificNotation[0].indexOf('.') === -1 ? '.' : '')//+ '0'.repeat(20); | |||
There was a problem hiding this comment.
Since the zeros are not appended anymore, there is no need to add a '.' after the mantissa if it is a round number
|
Is there any desire to merge this compact change ? |
|
I went ahead and publishes this breaking change by making a new package and naming it |
|
Oh I can't remember if there was something still wrong with this? Did all
the tests pass etc? Possibly I just forgot about it?
…On Wed, Mar 11, 2020, 04:20 Jake Verbaten ***@***.***> wrote:
I went ahead and publishes this breaking change by making a new package
and naming it charwise-compact (
https://github.com/Raynos/charwise-compact ).
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#9?email_source=notifications&email_token=AAB7KLXWURQBLJ4NYXGGUODRGZLDNA5CNFSM4E7REMN2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEOL3JTI#issuecomment-597144781>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAB7KLTHF4OJCV6OTWQT3WDRGZLDNANCNFSM4E7REMNQ>
.
|
|
From what i remember, everything was ok. |
|
I suspect you either forgot about this or did not want to do any breaking changes or migrations of the key space in leveldb. |
| @@ -6,6 +6,8 @@ | |||
| // We endpad mantissa with enough zero to exceed mantissa precision. | |||
There was a problem hiding this comment.
this comment is for the old code...
|
You have reviewed your own pull request. This is pretty meta. |
@PaulBlanche I realized that numbers are encoded with a huge amount of precision.
2 is encoded as
FE500M2.00000000000000000000... that seems unnecessary to me?I removed that, and noticed it failed on the close random number tests
so I tried just appending a number that we know is higher than 9, but only on negative numbers. This seems to work, tests pass.
cw.encode(-1)+'a' > cw.encode(-1.00000001)+'a'