Give rows where the all of the values in columns "a" and "b" are the same:

The order the keys are specified is used in the result but does not affect which rows match:

Give all row combinations where the all of the values in columns "a" and "b" in tab1 and "A" and "B" in tab2 are the same:

There are no combinations of rows that have values matching in both "a" and "B" and "b" and "A":

There is a single row combination that matches in column "a" from tab1 and "A" and "B" from tab2:

The combination is row 3 from tab1 and row 2 from tab2:

Include only rows where values in both columns "a" and "b" match:

This is the default and also equivalent to not requiring all rows from either tab1 or tab2:

Include an instance of all rows from both tab1 and tab2. Where there is not a match, values are filled in as missing:

Using {1,2} for the specification is equivalent:

With the specification {2,1}, the unmatched rows from tab2 are included before those of tab1:

Include an instance of all rows from tab1:

Using "Left" for the specification is equivalent since tab1 is the leftmost in the list:

Include an instance of all rows from tab2:

Using "Right" for the specification is equivalent since tab1 is the leftmost in the list:

Include only combinations of rows from three Tabular objects where the values of column "a" match:

This is the default and also equivalent to specifying no row requirement:

Include an instance of all rows from three Tabular objects:

Using {1,2,3} for the specification is equivalent:

Show the result of all possible specifications for the required rows for three Tabular objects:

When there are no matches, using "Outer" effectively stacks the Tabular objects:

This is equivalent to using Join:

Find combinations of rows where column "a" matches:

Using the key prefixes given with Association is necessary because there is a collision with column "b" where the values do not match:

When the match column names are given with specific Tabular objects, the matched columns are included with the prefixed names:

Include all the rows from two Tabular objects, merging row combinations where the values of column "b" match:

A mix of columns to match and merge and just to match may be used:

When the match column names are given with specific Tabular objects, the matched columns are included with the prefixed names and have missing elements where there is not a match:

Do a cross join (where any combination of rows is considered a match and included) of three Tabular objects:

In this case, the specification for required rows makes no difference, since all rows are included anyway:

Specify that the values from column "a" of tab1 or column "c" of tab2 are the same:

This is equivalent to using column names:

The function gets a single argument for each possible combination of rows; for row i from tab1 and row j from tab2, the argument is a list of Association objects, {Normal[tab1[i]],Normal[tab2[j]]}:

With Apply, the function gets two arguments and positional parameters may be used:

When the first argument is given as an association, the function gets an Association argument:

Find the combinations of rows where values in column "a" is the same and values in column "b" have the same length:

Find the combinations of rows where the parity of the values in column "a" is the opposite of the values in column "b":

Include all rows:

Include any combinations of rows that have matches in column "a" from tab1 and tab2, "c" from tab1 and column "b" from tab3:

Include all rows: