Unable to insert row in Spanner where column type is FLOAT64 and value is integer

[modeswitch]

My table has a FLOAT64 column, however some of my values happen to be integers (no fractional part). When I try to write these values, I get an error Invalid value for column X in table Y: Expected FLOAT64. If I add 0.0001 to each value before writing then the inserts are successful.

[jmuk]

@stephenplusplus -- is there a way to explicitly specify a float number here?

[stephenplusplus]

@vkedia - I'm experimenting with creating a spanner.double() method. The resulting API request sends the proto as:

{
  "insert": {
    "table": "Singers",
    "columns": [
      "SingerId"
    ],
    "values": [
      [
        {
          "numberValue": 8
        }
      ]
    ]
  }
}

An error is returned from the API:

Error: Row [8] in table Singers.

However, sending 8.1 as the value, the API request succeeds. Any idea if I'm making a simple mistake?

[vkedia]

Can you not set it to 8.0? I am not sure why 8 does not work. I will try to find out.

[stephenplusplus]

I think it's just a JavaScript thing: https://jsconsole.com/?8.0

[modeswitch]

Correct. If you want to test for float in JS you can do something like !(Number.isNaN(n) || Number.isInteger(n)).

Perhaps this is naive, but if you pass a numberValue to the Spanner API, shouldn't it be able to coerce an integer into a float?

[stephenplusplus]

@vkedia -- any preference how I should move forward with this?

[vkedia]

We do want people to be able to store a value like 8.0 in the database. I am surprised that sending 8 does not work. Can you do some magic to set it to 8.0 in the proto for now till I figure out whats going on with the backend?

[vkedia]

Oh I just realised this 8 vs 8.1 thing is a red herring. The error you got means that the row already exists in the table. Can you try with some other value which does not exist?