identify key variable for joining tables

I have two large tables (562000*2500) imported from a sql database with non-identical headers. I know these tables have 1 or 2 idetical column values which I want to use for joining the two tables. How is it possible to identify the potential key column for joining them?

[Geoff Hayes]

Johnny - how many columns/fields in each table? (Are you really saying there are 2500?)

[Johnny Birch]

Yes 2500 columns and > 500.000 rows.

[Geoff Hayes]

What are the data types of the columns that are common to both tables? Do you know if they represent the unique identifiers for the rows in the table?

[Johnny Birch]

the common datatype is strings. I am not sure what you mean by the second question, sorry

[Geoff Hayes]

Don't rows/records in tables usually have unique identifiers (the keys)? It's been a while since I've done any database development. But since the common data type is string, then this might not be a key. Are the strings unique for each record/row, or can multiple records have the same string?

[Guillaume]

So you have no a-priori knowledge which of the table columns may be the shared key?

And they're not the primary keys of the tables in the db (assuming your tables have primary keys) or indexed keys (which could reduce the search space)?

As Geoff points out primary keys are typically numerical integers. For strings, you'd use a hash to speed up search.

I've never heard of qxl databases.

[Johnny Birch]

You are right Guillaume. I have no a-priori knowledge about how the two tables are connected, except that they are. And I know they share one or two common column. They may be numerical, I am unsure since these two are large tables. It is from a sql database and I have no experience with sql.

[Johnny Birch]

@Geoff. I may be wrong. The common column may not be a string. I have not looked into all the values as the tables a quite large.

[Adam Danz]

So many of these questions (and more) could be clarified by including a mini example that represents your data.

[Guillaume]

With no a priori knowledge and no way to narrow it down, this sounds like a monumental task, if not impossible. In theory, you'd have to do up to 562000 ^ 3123750 comparisons to find the two columns with the most collisions.

If the two tables come from the same database, is there something in the database schema that could tell you the common column, i.e is there foreign keys?

[Geoff Hayes]

Johnny - how do you even know that there are common columns between the two tables? What has led you to believe that?