John Omernik created DRILL-3698:
-----------------------------------
Summary: Expose Show Files Command As SQL for sorting/filtering
Key: DRILL-3698
URL: https://issues.apache.org/jira/browse/DRILL-3698
Project: Apache Drill
Issue Type: Improvement
Components: SQL Parser
Affects Versions: Future
Environment: All
Reporter: John Omernik
Assignee: Aman Sinha
Fix For: Future
When using drill, I had a workspace setup, and I found myself using the show
files command often to find my directories etc. The thing is, the return of
show files is not ordered. And when looking at file system data there are many
possible ways to order the results for efficiency as a user.
Consider the ls command in unix; the ability to specify different sorting is
built in there. I checked out http://drill.apache.org/docs/show-files-command/
as well as tried the "obvious" show files order by name and that didn't work
nor did I see how I could in the documentation.
Based on a mailing list discussion there is no way to do that currently in
Drill, hence this JIRA I think just adding ORDER BY SQL methodology would be
perfect here, you have 8 fields (seen below) and ordering by any one of them,
or group of them, with ASC/DESC just like standard SQL order by would be a huge
win.
I suppose one could potentially ask for WHERE clause (filtering)too, and maybe
a select (which fieldsto display) however I am more concerned with the order,
but if I had to implement all there I could see examples below:
(All Three, select, where, and order) (I.e. after "Files" if the token isn't
WHERE or ORDER then check for the fields, if it's not a valid field list error)
SHOW FILES name, accessTime WHERE name like '%.csv' ORDER BY name;
(Where clause and order, note the token after FILES is WHERE)
SHOW FILES WHERE name like '%.csv' ORDER BY length ASC, name DESC;
(Only Order, ORDER Is the first token after FILES)
SHOW FILES ORDER BY length ASC, name DESC
I don't think we have to grant full SQL functionality here (i.e. aggregates),
just the ability to display various fields, filter on criteria, and ordering.
If you wanted to get fancy, I suppose you could take the table and make it a
full on table, i.e. take the results make it a quick inmemory table and then
utilize the whole drill stack on it. Lots of options. I just wanted to get
this down in an email/JIRA as it was something I found myself wishing I had
over and over during data exploration.
Fields Currently Returned:
|name| isDirectory|isFile|length|owner
group|permissions|accessTime|modificationTime|
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
