Aditya J.

Yeah I'm trying to change the codebase only

Hey, is there any way where we can put capping on no of records in a block like max records should be for ex 10000 in a single block.

im gonna share this, i do have one question, lets say one job is running for more than 8 hrs, can we check like what data is being processed exactly in that job?

Okay if these is the order of my definitions: FieldDefinition("linkedin_url", "string",MatchType.NULL_OR_BLANK, MatchType.EXACT), FieldDefinition("email", "string", MatchType.NULL_OR_BLANK, MatchType.EXACT), FieldDefinition("domain", "string", MatchType.NULL_OR_BLANK, MatchType.EXACT), FieldDefinition("full_name", "string",MatchType.NULL_OR_BLANK, MatchType.FUZZY), FieldDefinition("company_name", "string", MatchType.NULL_OR_BLANK, MatchType.FUZZY) These are null count: linkedin_url| email|domain|full_name| +------------+---------+------+---------+ | 516770|251455239| 0| 3395582| This is duplicate domain repeating 1.5 M times: domain |count | +---------------------+-------+ |linktr.ee |1566375| same for Full Name: full_name |count | +--------------+-------+ |NULL |3395582| | |18892 | |Amit Kumar |18096 | Is these creating issues

It won't ba same spread ofcourse Prod data is too large

Also my training data is clean and small ofcourse, prod data may be skewed

Okay let's consider the order okay , so If linkedin fails, it should go to full name fuzzy, then email which has 251M null records and followed on. But my point is how zingg choosing like which field to take consideration for blocking. So, if order, linkedin url should go, then what are conditions where it's not picking linkedin url for blocking

What do you mean by earlier fields are evaluated first?

And by deterministic, is there any rule in zingg that i can apply while prediction

If if 85 percent is null, it should not end in same block right