Hi,
> 1、why the start time is "2018-08-27 09:50:00" not "2018-08-27 09:53:00"?
When I define the window, the starttime is not set.
When no 'starttime' is defined, windows are aligned to the start of the
upper time magnitude. So, if your window is defined in minutes, it will be
aligned to the start of the hour and the first window will be the current
window.
In the case of a window of 5:00 minutes, it will be 9:00 - 9:05, 9:05-9:10,
9:10-9:15, ... 9:45-9:50, 9:50-9:55, ...
The first data point sets the internal 'event time clock' and the first
corresponding window for 9:53 is 9:50-9:55
Also, note that 'start time' is a very misleading name for that window
parameter. It's actually `starttimeoffset`. If you would specify
'startTime' to `1 minute`, then the windows will be 9:01-9:06, 9:06-9:11,
...
2、why the agg result of time "2018-08-27 09:53:00 " is not output when the
batch1 data is coming?
Yes. the resulting value is clearly there:
2018-08-27 09:50:00|2018-08-27 09:55:00| 2|
the two datapoints that fall in this window are summed and the result is 2.
What would were you expecting?
kr, Gerard.
On Mon, Aug 27, 2018 at 5:37 AM z...@zjdex.com <z...@zjdex.com> wrote:
> Hi Jungtaek Lim & Gerard Mass:
> Thanks very much.
> When I put three batch data like following :
>
> batch 0:
> 2018-08-27 09:53:00,1
> 2018-08-27 09:53:01,1
>
> batch 1:
> 2018-08-27 11:04:00,1
> 2018-08-27 11:04:01,1
>
> batch 2:
> 2018-08-27 11:17:00,1
> 2018-08-27 11:17:01,1
>
> the agg result of time "2018-08-27 09:53:00" is output like following:
> Batch: 2
> -------------------------------------------
> +-------------------+-------------------+--------+
> | start| end|sumvalue|
> +-------------------+-------------------+--------+
> |2018-08-27 09:50:00|2018-08-27 09:55:00| 2|
> +-------------------+-------------------+--------+
>
> For the result, I wonder to know:
> 1、why the start time is "2018-08-27 09:50:00" not "2018-08-27 09:53:00"?
> When I define the window, the starttime is not set.
> 2、why the agg result of time "2018-08-27 09:53:00 " is not output when
> the batch1 data is comming?
>
> Thanks a lot!
>
>
>
> ------------------------------
> z...@zjdex.com
>
>
> *From:* Jungtaek Lim <kabh...@gmail.com>
> *Date:* 2018-08-27 11:01
> *To:* z...@zjdex.com
> *CC:* Gerard Maas <gerard.m...@gmail.com>; user <user@spark.apache.org>;
> 王程浩 <w...@zjdex.com>
> *Subject:* Re: Re: About the question of Spark Structured Streaming
> window output
> You may want to add streaming listener to your query and see when/how
> watermark is updated. In short, watermark is calculated from previous batch
> and calculated value is applied to current batch. So you may think that the
> result is provided later than expected, maybe a batch.
>
> 2018년 8월 27일 (월) 오전 11:56, z...@zjdex.com <z...@zjdex.com>님이 작성:
>
>> Hi Gerard Mass:
>> Thanks a lot for your reply.
>> When i use "append" model, I send the following data:
>> 2018-08-27 09:53:00,1
>> 2018-08-27 09:53:01,1
>> The result (which has only schema, like the following) has received after
>> the batch is end. But when the time of window + watermark is up, there
>> is no result to output. Is there something I misss? Thanks in advance.
>>
>>
>>
>> ------------------------------
>> z...@zjdex.com
>>
>>
>> *From:* Gerard Maas <gerard.m...@gmail.com>
>> *Date:* 2018-08-27 05:00
>> *To:* zrc <z...@zjdex.com>
>> *CC:* spark users <user@spark.apache.org>; wch <w...@zjdex.com>
>> *Subject:* Re: About the question of Spark Structured Streaming window
>> output
>>
>> Hi,
>>
>> When you use a window in Append mode, you need to wait for the end of the
>> window + watermark to see the final record from the "append" mode.
>> This is your query over time. Note the timestamp at the right side of the
>> cell and the data present in it.
>>
>> val windowedCounts = dataSrc
>> .withWatermark("timestamp", "1 minutes")
>> .groupBy(window($"timestamp", "5 minutes"))
>> .agg(sum("value") as "sumvalue")
>> .select("window.start", "window.end","sumvalue")
>>
>>
>> [image: image.png]
>>
>> Going back to your questions:
>> 1、when I set the append output model, I send inputdata, but there is no
>> result to output. How to use append model in window aggreate case ?
>> Wait for the window + watermark to expire and you'll see the append
>> record output
>>
>> 2、when I set the update output model, I send inputdata, the result is
>> output every batch .But I want output the result only once when window is
>> end. How can I do?
>> Use `append` mode.
>>
>> kr, Gerard.
>>
>> On Thu, Aug 23, 2018 at 4:31 AM z...@zjdex.com <z...@zjdex.com> wrote:
>>
>>> Hi :
>>> I have some questions about spark structured streaming window output
>>> in spark 2.3.1. I write the application code as following:
>>>
>>> case class DataType(time:Timestamp, value:Long) {}
>>>
>>> val spark = SparkSession
>>> .builder
>>> .appName("StructuredNetworkWordCount")
>>> .master("local[1]")
>>> .getOrCreate()
>>>
>>> import spark.implicits._
>>> // Create DataFrame representing the stream of input lines from
>>> connection to localhost:9999
>>> val lines = spark.readStream
>>> .format("socket")
>>> .option("host", "localhost")
>>> .option("port", 9999)
>>> .load()
>>>
>>> val words = lines.as[String].map(l => {
>>> var tmp = l.split(",")
>>> DataType(Timestamp.valueOf(tmp(0)), tmp(1).toLong)
>>> }).as[DataType]
>>>
>>> val windowedCounts = words
>>> .withWatermark("time", "1 minutes")
>>> .groupBy(window($"time", "5 minutes"))
>>> .agg(sum("value") as "sumvalue")
>>> .select("window.start", "window.end","sumvalue")
>>>
>>> val query = windowedCounts.writeStream
>>> .outputMode("update")
>>> .format("console")
>>> .trigger(Trigger.ProcessingTime("10 seconds"))
>>> .start()
>>>
>>> query.awaitTermination()
>>>
>>> the input data format is :
>>> 2018-08-20 12:01:00,1
>>> 2018-08-20 12:02:01,1
>>>
>>> My questions are:
>>> 1、when I set the append output model, I send inputdata, but there is no
>>> result to output. How to use append model in window aggreate case ?
>>> 2、when I set the update output model, I send inputdata, the result is
>>> output every batch .But I want output the result only once when window is
>>> end. How can I do?
>>>
>>> Thanks in advance!
>>>
>>>
>>> ------------------------------
>>> z...@zjdex.com
>>>
>>
