从ISO 8601日期字符串转换为BigQuery时间戳时出错

Question

I have an ISO 8601 date string in the following format:

String myIsoDateString = "2019-02-27T23:00:00.000Z"

I need to use the date string as part of a query I'm running in BigQuery. I'm trying to use the com.google.cloud.bigquery.QueryParameterValue class to convert it to a QueryParameterValue , with a type of timestamp , like this:

QueryParameterValue.timestamp(myIsoDateString)

This gives me an error:

java.lang.IllegalArgumentException: Invalid format: "2019-02-27T23:00:00.000Z" is malformed at "T23:00:00.000Z"

The inline help in Eclipse for the timestamp method states that it:

Creates a QueryParameterValue object with a type of TIMESTAMP. Must be in the format"yyyy-MM-dd HH:mm:ss.SSSSSSZZ", e.g. "2014-08-19 12:41:35.220000+00:00".

How do I convert myIsoDateString to the required format? Is there a better method I can use that will handle converting from an ISO 8601 string to a timestamp in BigQuery?

Answer 1

I would expect the answer the answer by Felipe Hoffa to work and give you what you want: replace the T with a space. Here are a couple of other suggestions.

Be explicit: While the documentation gives a quite clear example of an expected string, it's not clear from the error message nor from the documentation (link below, but you've got it in Eclipse already) whether QueryParameterValue.timestamp accepts the Z as an offset. If we want to be sure and also be more explicit about supplying the format asked for (always nice for those managing your code after you):

    DateTimeFormatter timestampFormatter
            = DateTimeFormatter.ofPattern("uuuu-MM-dd HH:mm:ss.SSSSSSxxx");
    String myIsoDateString = "2019-02-27T23:00:00.000Z";
    OffsetDateTime dateTime = OffsetDateTime.parse(myIsoDateString);
    QueryParameterValue.timestamp(dateTime.format(timestampFormatter));

This passes the string 2019-02-27 23:00:00.000000+00:00 to timestamp .

Pass a long: timestamp comes in an overloaded version taking a Long argument. Again it's not documented what the long value should be. Edit: Your own answer revealed that timestamp(Long) requires microseconds since the epoch. The fully correct way to convert without any risk of unnecessary loss of precision is:

    String myIsoDateString = "2019-02-27T23:00:00.000Z";
    Instant dateTime = Instant.parse(myIsoDateString);
    long microsecondsSinceEpoch = TimeUnit.SECONDS.toMicros(dateTime.getEpochSecond())
            + TimeUnit.NANOSECONDS.toMicros(dateTime.getNano());
    QueryParameterValue.timestamp(microsecondsSinceEpoch);

A value of 1 551 308 400 000 000 is passed to timestamp .

Disclaimer: I'm not a BigQuery user.

Links

• QueryParameterValue documentation
• Oracle tutorial: Date Time

Answer 2

The error says this is not the right format:

 "2019-02-27T23:00:00.000Z"

But this one would be:

 "yyyy-MM-dd HH:mm:ss.SSSSSSZZ"

Try replacing the T between date and time with a .

Answer 3

ISO 8601 versus SQL

The ISO 8601 standard uses a T in the middle to separate the year-month-day portion from the hour-minute-second portion.

The SQL style uses a SPACE in the middle rather than a T .

Replace.

String input = "2019-02-27T23:00:00.000Z".replace( "T" , " " ) ;
QueryParameterValue.timestamp( input ) ;

Ditto for going the other direction. The java.time classes use the ISO 8601 formats by default when parsing/generating text.

String input = "2019-02-27 23:00:00.000Z".replace( " " , "T" ) ;
Instant instant = Instant.parse( input ) ;
ZonedDateTime zdt = instant.atZone( ZoneId.of( "Pacific/Auckland" ) ) ;

Answer 4

This works:

import org.joda.time.DateTime;
import org.joda.time.DateTimeZone;

Convert myIsoDateString to microseconds and store it as a long :

long myDateInMicroseconds = DateTime.parse(myIsoDateString).toDateTime(DateTimeZone.UTC).getMillis() * 1000;

Then pass it as the argument to QueryParameterValue.timestamp:

QueryParameterValue.timestamp(myDateInMicroseconds)

Answer 5

To anyone who is just trying to parse ISO 8601 in BigQuery (this post is the first Google result), try this:

SELECT
PARSE_TIMESTAMP('%Y-%m-%dT%H:%M:%E*SZ',  '2018-10-12T13:22:27.120Z')