[
https://issues.apache.org/jira/browse/DRILL-3435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14610646#comment-14610646
]
Andy Pernsteiner commented on DRILL-3435:
-----------------------------------------
Maybe I'm missing something here, but this is a relatively specific case, and
in fact, my queries against other columns in this JSON file do NOT require any
table aliasing, because this is not a nested JSON doc (at least, not the fields
I'm going after) EG:
{code}
: jdbc:drill:> select id from profiles limit 3;
+-----------------------------------------------------------------+
| id |
+-----------------------------------------------------------------+
| {"part1":"3069735710943609483","part2":"-1130875713383276112"} |
| {"part1":"3069511572705759702","part2":"-8164107516750810817"} |
| {"part1":"3057076082656734793","part2":"-4260511657052163769"} |
+-----------------------------------------------------------------+
3 rows selected (2.839 seconds)
{code}
So that works, however running w/ `user` does not:
{code}
0: jdbc:drill:> select `user` from profiles limit 3;
+---------------+
| user |
+---------------+
| apernsteiner |
| apernsteiner |
| apernsteiner |
+---------------+
3 rows selected (2.141 seconds)
{code}
Also, I'm not sure what you mean by your statement :
{quote}
..'user' is recognized as a function in this case, and the query fails..
{quote}
What do you mean by the query fails? It doesn't actually fail, it returns data
which would be confusing to someone who ran a query against another column in
this file/table, where they got the actual data, and not the result of a
special function.
> Some reserved-keywords require table aliasing
> ---------------------------------------------
>
> Key: DRILL-3435
> URL: https://issues.apache.org/jira/browse/DRILL-3435
> Project: Apache Drill
> Issue Type: Bug
> Components: Documentation
> Affects Versions: 1.0.0
> Reporter: Andy Pernsteiner
> Assignee: Bridget Bevens
> Priority: Minor
> Labels: documentation
>
> Not only does drill have a number of reserved keywords that require
> backticking (``), there also appear to be some reserved words that require
> extra care, using table aliases to be able to perform queries. One that
> we've found so far is 'user' . EG, consider the following scenario:
> bq. /usr/bin/sqlline -u jdbc:drill: -n root
> then:
> {code} select user from
> `profiles/2aa32e9e-bdae-8949-8461-c14dafe63ee0.sys.drill` ;
> +-------+
> | user |
> +-------+
> | root |
> +-------+
> {code}
> But the actual file in question has the 'user' as a different user:
> {code} cat 2aa32e9e-bdae-8949-8461-c14dafe63ee0.sys.drill|egrep -o
> 'user\":\"[a-z]+\"'
> user":"apernsteiner"
> {code}
> The workaround is to alias the table (t) and prefix the 'user' column in the
> resultset w/ the table alias :
> {code}
> 0: jdbc:drill:> select t.`user` from
> `profiles/2aa32e9e-bdae-8949-8461-c14dafe63ee0.sys.drill` t ;
> +---------+
> | user |
> +---------+
> | apernsteiner |
> +---------+
> {code}
> @jinfeng gave the following explanation on the user@ list:
> {quote}
> 'user' is a SQL reserved word.
> When it's used alone, it is a system function, just like CURRENT_USER. See
> http://calcite.incubator.apache.org/docs/reference.html (System functions
> section).
> When 'user' is qualified with a table alias, it becomes a column
> identifier.
> {quote}
> The drill documentation @ https://drill.apache.org/docs/reserved-keywords/
> merely says to use backticks (``), not to do any table aliasing. For those
> who have columns named 'user', this may be misleading...
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
