Add databricks direct unload file containing complex map key

[njaremko]

Required for 7769

[alamb]

Thank you @njaremko 🙏

Could you please add some comments about what this file contains and how you created it?

Ideally you could follow the model of https://github.com/apache/parquet-testing/blob/master/data/README.md

Also I noticed the file is 8KB -- given how widely this repo is cloned / copied is there any way to make the example file smaller?

[njaremko]

I've removed the unneeded columns, and it's 2kb now. I've also updated the readme

[njaremko]

What are the next steps to get this merged?

[alamb]

Thank you @njaremko . This looks good to me

I dumped schema and layout ofdata/complex_map_key.parquet and it looks good to me

Note that I am not a committer on parquet so I can not commit this PR. Perhaps @wgtmac or @emkornfield could take a look.

Also FYI I don't think we need to gate the fix in arrow-rs on this PR. I will comment on apache/arrow-rs#7769 as well

parquet-layout data/complex_map_key.parquet
required group field_id=-1 spark_schema {
  required group field_id=-1 map_nested (Map) {
    repeated group field_id=-1 key_value {
      required binary field_id=-1 key (String);
      required group field_id=-1 value (Map) {
        repeated group field_id=-1 key_value {
          required binary field_id=-1 key (String);
          required binary field_id=-1 value (String);
        }
      }
    }
  }
  required group field_id=-1 map_nested_array (Map) {
    repeated group field_id=-1 key_value {
      required group field_id=-1 key (List) {
        repeated group field_id=-1 list {
          required int32 field_id=-1 element;
        }
      }
      required group field_id=-1 value (Map) {
        repeated group field_id=-1 key_value {
          required binary field_id=-1 key (String);
          required int32 field_id=-1 value;
        }
      }
    }
  }
}

and

File Name: data/complex_map_key.parquet
Version: 1.0
Created By: parquet-mr version 1.12.3-databricks-0002 (build 2484a95dbe16a0023e3eb29c201f99ff9ea771ee)
Total rows: 1
Number of RowGroups: 1
Number of Real Columns: 2
Number of Columns: 6
Number of Selected Columns: 6
Column 0: map_nested.key_value.key (BYTE_ARRAY / String / UTF8)
Column 1: map_nested.key_value.value.key_value.key (BYTE_ARRAY / String / UTF8)
Column 2: map_nested.key_value.value.key_value.value (BYTE_ARRAY / String / UTF8)
Column 3: map_nested_array.key_value.key.list.element (INT32)
Column 4: map_nested_array.key_value.value.key_value.key (BYTE_ARRAY / String / UTF8)
Column 5: map_nested_array.key_value.value.key_value.value (INT32)
--- Row Group: 0 ---
--- Total Bytes: 256 ---
--- Total Compressed Bytes: 266 ---
--- Rows: 1 ---
Column 0
  Values: 1, Null Values: 0, Distinct Values: 0
  Max: a, Min: a
  Compression: SNAPPY, Encodings: PLAIN
  Uncompressed Size: 40, Compressed Size: 42
Column 1
  Values: 1, Null Values: 0, Distinct Values: 0
  Max: b, Min: b
  Compression: SNAPPY, Encodings: PLAIN
  Uncompressed Size: 42, Compressed Size: 44
Column 2
  Values: 1, Null Values: 0, Distinct Values: 0
  Max: c, Min: c
  Compression: SNAPPY, Encodings: PLAIN
  Uncompressed Size: 42, Compressed Size: 44
Column 3
  Values: 2, Null Values: 0, Distinct Values: 0
  Max: 2, Min: 1
  Compression: SNAPPY, Encodings: PLAIN
  Uncompressed Size: 45, Compressed Size: 47
Column 4
  Values: 1, Null Values: 0, Distinct Values: 0
  Max: green, Min: green
  Compression: SNAPPY, Encodings: PLAIN
  Uncompressed Size: 46, Compressed Size: 46
Column 5
  Values: 1, Null Values: 0, Distinct Values: 0
  Max: 5, Min: 5
  Compression: SNAPPY, Encodings: PLAIN
  Uncompressed Size: 41, Compressed Size: 43
--- Values ---
key                           |key                           |value                         |element                       |key                           |value                         |
a                             |b                             |c                             |1                             |green                         |5                             |
2                             |

[wgtmac]

Do we actually need the map_nested field if we just want to add complex key/value types?

[wgtmac]

Is it worth describing the exact schema of the file at this line?