[ZEPPELIN-2111] Table display is losing precision on large integers, even if it is cast as a string

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Cannot Reproduce
• Affects Version/s: 0.6.2
• Component/s: front-end
• Labels: display
• Assignee: Hoon Park
• Reporter: Anton
• Created: 14/Feb/17 03:33
• Updated: 19/Apr/17 04:59
• Resolved: 19/Apr/17 04:52

Description

When displaying large integers in the table display, it rounds as if it was double precision.

This does not happen with spark. See below for an example:

// This code produces the correct results
spark.sql("""select 2010090230001410131
    ,   cast("2010090230001410131" as string)
    ,   cast("S2010090230001410131" as string)
    ,   cast(2010090230001410131 as bigint)""").show

+-------------------+-----------------------------------+------------------------------------+-----------------------------------+
|2010090230001410131|CAST(2010090230001410131 AS STRING)|CAST(S2010090230001410131 AS STRING)|CAST(2010090230001410131 AS BIGINT)|
+-------------------+-----------------------------------+------------------------------------+-----------------------------------+
|2010090230001410131|                2010090230001410131|                S2010090230001410131|                2010090230001410131|
+-------------------+-----------------------------------+------------------------------------+-----------------------------------+

However if we display the dataframe above using Zeppelin's table mode, we get a different result.

%sql
select  2010090230001410131
    ,   cast("2010090230001410131" as string)
    ,   cast("S2010090230001410131" as string)
    ,   cast(2010090230001410131 as bigint)

Issue Links

• relates to: ZEPPELIN-1915 [Umbrella] Improve built-in visualizations