You can can eliminate the header row with a predicate. Also Drill will try to
guess the data type of columns, but sometimes it may be wrong. In this case the
header row can confuse Drill as to the data type of the column as the first
column has a string, and the next ones have a number. In such
That works with the file that has no row header, Sudhakar.
On Sun, Feb 8, 2015 at 10:34 PM, Sudhakar Thota wrote:
> May be you have to cast it to integer.
>
> Sudhakar Thota
> Sent from my iPhone
>
> > On Feb 8, 2015, at 6:54 PM, Minnow Noir wrote:
> >
> > The error actually happens when I remo
May be you have to cast it to integer.
Sudhakar Thota
Sent from my iPhone
> On Feb 8, 2015, at 6:54 PM, Minnow Noir wrote:
>
> The error actually happens when I remove the header row from the file.
>
> cat test2.csv
> "Ed",100
> "Pete",200
> "Ed",100
> "Pete",400
>
>
> 0: jdbc:drill:zk=local
The error actually happens when I remove the header row from the file.
cat test2.csv
"Ed",100
"Pete",200
"Ed",100
"Pete",400
0: jdbc:drill:zk=local> select columns[0], columns[1] from
dfs.`/data/test2.csv`;
+++
| EXPR$0 | EXPR$1 |
+++
| "Ed
You need to filter out the header line on the CSV file as you are trying to sum
a string in column1.
—Andries
On Feb 8, 2015, at 6:12 PM, Minnow Noir wrote:
> I'm trying to perform a basic query in order to learn Drill, but getting an
> the error message in the subject line.
>
> I created a
I'm trying to perform a basic query in order to learn Drill, but getting an
the error message in the subject line.
I created a dead simple CSV file on disk. Note that Sales values are not
quoted.
cat test.csv
Employee,Sales
Ed,100
Pete,200
Ed,100
Pete,400
When I query it without performing a s
On the HDP try to connect Drill to the ZK and then use the Squirrel JDBC
instructions.
Basically edit the /conf/drill-overrride.conf file
Take note of the cluster-id in the conf file (as you will use that to connect
to).
Then edit the zk.connect line for the node name with ZK on it. Also with
You can run in embedded mode within Squirrel but you would need to use the
same classpath that the drillbit startup creates for your driver rather
than using the jdbc-driver jar (the jdbc-driver jar doesn't support
embedded mode). Once started in embedded, the web interface will be
available.
On
Thanks Andries (and also, Aditya). I will try that.
On Sun, Feb 8, 2015 at 5:50 PM, Andries Engelbrecht <
aengelbre...@maprtech.com> wrote:
> CTAS (Create Table As) is the way to get a CSV output. Make sure you point
> to the correct directory and storage plugin configuration.
>
> Also to the be
Thanks, Kristine. Those are some helpful pointers.
On Sun, Feb 8, 2015 at 4:50 PM, Kristine Hahn wrote:
> I sent the attachment to you in a private email. I think the problem is
> that you need to specify the workspace, as shown below. Also included below
> is a portion of the steps for queryin
Embedded mode, since I'm just experimenting with it. I had it on another
HDP sandbox VM a couple of months ago, and IIRC, was able to access the web
UI from the host (unlike right now), but I never tried accessing it via a
thick client from the host.
I'm using the HDP 2.2 sandbox VM. ZK is on i
CTAS (Create Table As) is the way to get a CSV output. Make sure you point to
the correct directory and storage plugin configuration.
Also to the best way to ensure you get the CSV output file is set the session
setting for output to CSV.
I.e. alter session set `store.format` = ‘csv’;
As an F
Are you running Drill in embedded mode or using the Sandbox VM?
I don’t believe you can connect to a VM running Drill in Embedded mode from the
host using SquirreL with JDBC. JDBC connects via ZooKeeper.
It may be best to consider using the Drill SandBox if you would like to use
SQuirreL,
What
I sent the attachment to you in a private email. I think the problem is
that you need to specify the workspace, as shown below. Also included below
is a portion of the steps for querying multiple files in a directory.
Connect to and Query a File
When querying the same data source repeatedly, avoi
Thanks for confirming that I should be able to query multiple files,
Kristine. Your attachment didn't make it through. Was my syntax off, or do
I have another problem?
On Sun, Feb 8, 2015 at 4:00 PM, Kristine Hahn wrote:
> Minnow, a single-query of multiple TSV files works for me. Here's how:
>
Minnow, a single-query of multiple TSV files works for me. Here's how:
"Query Multiple Files in a Directory" at the end of the attached PDF.
Please overlook the code formatting and some links that didn't copy/paste
well.
On Sun, Feb 8, 2015 at 12:35 PM, Minnow Noir wrote:
> "Failure while runnin
"Failure while running fragment."
Here's an example doing it the way you suggest. (Last 100 lines of
sqlline.log are also attached.) It sounds like you think the multi-file
query should actually work?
Thanks
0: jdbc:drill:zk=local> select * from dfs.`/data` limit 5;
++
| column
Please see
https://cwiki.apache.org/confluence/display/DRILL/CREATE+TABLE+AS+%28CTAS%29+Command
On Sun, Feb 8, 2015 at 10:47 AM, Minnow Noir wrote:
> I'm trying to extract query results as CSV for ingestion into another
> tool. There's no obvious way in sqlline, the web UI, or the docs to
> ext
I am trying to access a drill bit running in a VM from Squirrel on my
desktop. I'm unsure of the Squirrel alias syntax, and having an issue
troubleshooting because I can't access the web interface on 8047 in the
VM. I'm trying this on a VM before worrying about deploying it into a
real cluster.
What is the error that you are seeing?
Can you simply point it to the directory (without *.csv) to see if it helps.
On Sun, Feb 8, 2015 at 10:33 AM, Minnow Noir wrote:
> I'm trying to do ad-hoc exploration/analysis over multiple files without
> having to concatenate them. New files show up on
I'm trying to extract query results as CSV for ingestion into another
tool. There's no obvious way in sqlline, the web UI, or the docs to
extract query results as CSV (or other formats).
sqlline ./test.sql > output.csv doesn't work. Is there a way to pass
sqlline a query file and then dump the ou
I'm trying to do ad-hoc exploration/analysis over multiple files without
having to concatenate them. New files show up on a regular basis, and
creating large, redundant concatenated files seems inelegant for data
exploration. I've tried the obvious (... from dfs.`/dir/*.csv` but that
only returns
22 matches
Mail list logo
