We are performing tests for an aggregator application, and are observing some differences on how data is being displayed for certain datatypes (double and decimal so far) between Cassandra and YugabyteDB.
For transaction table, market_value column with double datatype – our Spark application writes data (original value is converted into BigDecimal in Java) and the RDD prints value as 1000000.01. Once this row is written into the database, Cassandra shows this value as expected, whereas in YB, value in YCQLSH it shows up as 1e+06. Even our API that reads the data gets it as 1e+06 .
YB version – 2.1.6.2
YCQL java driver – 4.6.0-yb-6
Any thoughts or suggestions on what could be possible reasons for this value to be displayed differently?
This is our schema:
CREATE TABLE IF NOT EXISTS transaction (
account_id int,
transaction_id text,
quantity double,
price double,
market_value double,
PRIMARY KEY ((account_id), transaction_id)
) WITH CLUSTERING ORDER BY (transaction_id ASC);
Those non-integer types like DECIMAL, DOUBLE, FLOAT are inexact numbers. DOUBLE and FLOAT have limits on their precision (8 vs 4 bytes) and DECIMAL is an arbitrary precision type that does not have limits on the amount of precision. You can CAST(field as TEXT) to show it “expanded” to 1000000 instead of 1E+06.
DOUBLE has a limit of 15 digits after the decimal point and FLOAT has a limit 7 digits of precision.