is there way to convert scientific notation numbers to fixed length float

[jgoodall]

I have numbers like this 1.26126e-06 in JSON that I would like to convert to 'normal' looking floats (e.g 0.00000126126). Is there an equivalent to printf or JavaScript's toFixed()?

Or is there some other way to set the precision.

[MarSoft]

At least something like an option to convert float to int dropping the remainder. Then we could at least mock mynumber.toFixed(2) with (mynumber*100 | toInteger) / 100.

[sjackman]

I'd like a command line argument to specify the number of digits. FWIW you can do this using Miller (mlr):

$ mlr --json format-values -f '%.9f' <<<'{"x": 1.26126e-6}'
{ "x": 0.000001261 }

[honzajde]

@sjackman mlr --json format-values ... will not handle JSON array correctly (members are pulled out)

[sjackman]

Yeah, Miller handles only tabular JSON files AFAIK.

[jolmg]

@MarSoft This works right now: .*100 | floor | ./100.

[sjackman]

$ jq '.*100 | floor | ./100' <<<'{"hi": "Hello, world!", "frac": 123.456}'
jq: error (at <stdin>:1): object ({"hi":"Hell...) and number (100) cannot be multiplied

[itchyny]

@sjackman jq '(.. | numbers) |= (.*100 | floor | ./100)' <<<'{"hi": "Hello, world!", "frac": 123.456789}'

[sjackman]

That works! Thanks, @itchyny! As a feature request, round could optionally take an accuracy parameter such as round(0.01).

[jolmg]

@sjackman You can substitute floor with round in my suggestion to get the effect you want:

$ jq '.*100 | round | ./100' <<< 123.456789
123.46
$ jq '(.. | numbers) |= (.*100 | round | ./100)' <<<'{"hi": "Hello, world!", "frac": 123.456789}'
{
  "hi": "Hello, world!",
  "frac": 123.46
}

And also define the function you seek:

$ jq 'def round(p): .*(1/p) | round | ./(1/p); round(0.01)' <<< 123.456789
123.46

[ghost]

@sjackman You can substitute floor with round in my suggestion to get the effect you want:

$ jq '.*100 | round | ./100' <<< 123.456789
123.46
$ jq '(.. | numbers) |= (.*100 | round | ./100)' <<<'{"hi": "Hello, world!", "frac": 123.456789}'
{
  "hi": "Hello, world!",
  "frac": 123.46
}

And also define the function you seek:

$ jq 'def round(p): .*(1/p) | round | ./(1/p); round(0.01)' <<< 123.456789
123.46

can this be adjusted to work with larger numbers?

$ jq 'def round(p): .*(1/p) | round | ./(1/p); round(0.00001)' <<< 0.00004948914415535700
5.000000000000001e-05

[jolmg]

@arno01 You can simplify the expressions I used. I used those because they seemed clearer to me, but if you limit the operations used, that decreases chances of floating-point rounding errors:

$ jq 'def round(p): ./p | round | .*p; round(0.00001)' <<< 0.00004948914415535700
5e-05

[ghost]

@arno01 You can simplify the expressions I used. I used those because they seemed clearer to me, but if you limit the operations used, that decreases chances of floating-point rounding errors:

$ jq 'def round(p): ./p | round | .*p; round(0.00001)' <<< 0.00004948914415535700
5e-05

I'd love to get rid of scientific notation though.. if it could turn 0.00004948914415535700 into 0.0000495 that'd be neat :)

[jolmg]

@arno01 Adding that to jq is what this issue is about. In the mean time, converting that scientific notation would have be done outside. For example, like this:

$ printf '%.7f\n' "$(jq 'def round(p): ./p | round | .*p; round(0.0000001)' <<< 0.00004948914415535700)"
0.0000495

[adrian-gierakowski]

needed here as well

[loskael]

$ echo '{}' | jq '0.1  + 0.2 | .*100 | floor | ./100'
0.3

The expected result should be 0.30