Search code examples
sqlapache-drill

Data spills out of table in Apache Drill

I've been using Apache Drill lately for checking my raw data. It's been going great, however I noticed that it doesn't pretty print or wrap when showing the data in a table. For example:

    +----+-------------+
    | ID | DESCRIPTION |
    +----+-------------+
    | 25840 | Lorem ipsum dolor sit amet |
    | 27127 | Vivamus pharetra at purus et porta |
    | 27128 | Maecenas sagittis lectus urna, at facilisis arcu facilisis ut |
    | 27753 | Integer id lobortis nunc. Donec nec lacus feugiat, 
tempor tellus nec, mattis sem |
    | 28034 | Praesent vulputate id neque non aliquet |
    +----+-------------+

As opposed to looking like this:

+--------+-----------------------------+
|   ID   |         DESCRIPTION         |
+--------+-----------------------------+
| 25840  | Lorem ipsum dolor sit amet  |
+--------+-----------------------------+

Not the end of the world, however it makes viewing the data kind of frustrating. Notice how the row with the ID of 27753 is too long for my terminal, so it breaks to a new line. Instead of just expanding the row height, it expands every single column length past the header size, and then makes all of them do it, even though they'd fit within the terminal size. I imagine it would look like this:

    +-------+----------------------------------------+
    |   ID  |            DESCRIPTION                 |
    +----+-------------------------------------------+
    | 25840 | Lorem ipsum dolor sit amet             |
    | 27127 | Vivamus pharetra at purus et porta     |
    | 27753 | Integer id lobortis nunc. Donec nec    |
    |       |  lacus feugiat, tempor tellus nec,     |
    |       |  mattis sem                            |
    | 28034 | Praesent vulputate id neque non aliquet|
    +-------+----------------------------------------+

There might not be a way to do this yet, but before I submitted a feature request on JIRA, I figured I'd ask.

Solution

  • The "pretty printing" is done by SqlLine (https://github.com/julianhyde/sqlline), based on the schema provided by the JDBC. Since Drill is "schema free", it discovers the schema(s) as the data is processed by the query, and returns the discovered schema(s) with the data. It is possible that in your test, the first schema was somehow returned without any data, then the schema changed to "wider" data types.