Team,

Do you know if it's possible to use multiple CRON expressions in a single definition? CRON provides a lot of flexibility within each given time unit however it would be nice to combine them too

Thanks

For an Android app, I had a need to parse a Cron expression and determine the previous/next execution date based on this expression. I came across this library but then realized it relies on JSR310 which was introduced in Java 8 and does not work with Android.

I forked the library and was able to make it work on Android in 2 steps:

Step 1: Make it work with Java 7

1. Use threetenbp which is a backport of JSR310
2. Search-and-replace all occurrences of java.time with org.threeten.bp
3. Use retrolambda to be able to use java.util.function.Function and the lambdas in pre-Java 8

With this, I was able to successfully run all cron-util tests with a target of Java 7

Step 2: Make it work on Android

Here I started off with some tests for my own use case (parse a Cron expression and get the next/previous execution times), but I made them Android tests (i.e., run on an emulator). For this. I had to make some more changes in my app/tests

1. Add ThreeTenABP which is an Android extension to threetenbp.
2. Initialize ThreeTenABP in my test like this. If you use it in an app, you will still have to init it in your Application sub-class.
3. Add a dependency on an implementation of SL4J. I used this one but any which works for you will do.

compile 'eu.lp0.slf4j:slf4j-android:1.7.22-0'

With these changes, I was able to run my app-specific tests successfully on Android.

Next steps

I left it at that because I am not sure how the maintainers would like to proceed with this. There are open questions like

• Do you want to support Android in the first place?
• If so, would you like to use JSR310 for Java8 targets and use the backport for older targets? Basically generate multiple versions of the library from the same code-base with multiple targets
• Would you rather do some runtime shenanigans to detect if JSR310 is on the classpath and use that and fall back to the backport otherwise (don't even know if this is possible!)
• How would you like to approach the tests? For the Android target, It makes sense to run all the existing (408, at the time of this writing) tests on an emulator to make sure everything is compatible.

Utilizing this in Java 8 requires either, that the project uses Guava and ThreeTenBP replacements in the entire code base, or significant boilerplate to map between Guava/ThreeTenBP and Java 8.

Would a separate release branch for Java 8 be feasible?

I could drive with the migration in that case.

For the cron expression 0/30 * * * * *, with a custom definition of:

// Unix Crontab with seconds allowed CronDefinition cronDefinition = CronDefinitionBuilder.defineCron() .withSeconds().and() .withMinutes().and() .withHours().and() .withDayOfMonth().and() .withMonth().and() .withDayOfWeek().withValidRange(0, 7).withMondayDoWValue(1).withIntMapping(7, 0).and() .instance();

Performing a nextExecution call when the input DateTime value is between 30 and 59 seconds of the minute causes the method to return with a DateTime at the previous minute instead of the next minute.

For example:

nextExecution('2015-08-28 12:05:44') -> 2015-08-28 12:05:00 nextExecution('2015-08-28 12:05:14') -> 2015-08-28 12:05:30

Note that I've also tried this with '*/30 * * * * *' and got the same results.

Unit tests:

@Test
public void testCronExpressionAfterHalf() {

    CronDefinition cronDefinition = CronDefinitionBuilder.defineCron()
            .withSeconds().and()
            .withMinutes().and()
            .withHours().and()
            .withDayOfMonth().and()
            .withMonth().and()
            .withDayOfWeek().withValidRange(0, 7).withMondayDoWValue(1).withIntMapping(7, 0).and()
            .instance();

    CronParser parser = new CronParser(cronDefinition);
    Cron cron = parser.parse("*/30 * * * * *");

    MutableDateTime mutableDateTime = new MutableDateTime();
    mutableDateTime.setDateTime(2015, 8, 28, 12, 5, 44, 0);

    DateTime startDateTime = mutableDateTime.toDateTime();

    mutableDateTime = new MutableDateTime();
    mutableDateTime.setDateTime(2015, 8, 28, 12, 6, 0, 0);

    DateTime expectedDateTime = mutableDateTime.toDateTime();

    ExecutionTime executionTime = ExecutionTime.forCron(cron);

    DateTime nextExecutionDateTime = executionTime.nextExecution(startDateTime);
    MatcherAssert.assertThat(nextExecutionDateTime, Matchers.equalTo(expectedDateTime));
}

@Test
public void testCronExpressionBeforeHalf() {

    CronDefinition cronDefinition = CronDefinitionBuilder.defineCron()
            .withSeconds().and()
            .withMinutes().and()
            .withHours().and()
            .withDayOfMonth().and()
            .withMonth().and()
            .withDayOfWeek().withValidRange(0, 7).withMondayDoWValue(1).withIntMapping(7, 0).and()
            .instance();

    CronParser parser = new CronParser(cronDefinition);
    Cron cron = parser.parse("0/30 * * * * *");

    MutableDateTime mutableDateTime = new MutableDateTime();
    mutableDateTime.setDateTime(2015, 8, 28, 12, 5, 14, 0);

    DateTime startDateTime = mutableDateTime.toDateTime();

    mutableDateTime = new MutableDateTime();
    mutableDateTime.setDateTime(2015, 8, 28, 12, 5, 30, 0);

    DateTime expectedDateTime = mutableDateTime.toDateTime();

    ExecutionTime executionTime = ExecutionTime.forCron(cron);

    DateTime nextExecutionDateTime = executionTime.nextExecution(startDateTime);
    MatcherAssert.assertThat(nextExecutionDateTime, Matchers.equalTo(expectedDateTime));
}

The declared Guava dependency by the POM is too restrictive. The interval of compatible versions could be opened up to the current version, i.e. 23.4-jre.

Now that the sepcific version is resolved by the org.apache.felix plugin to [18.0,19) and injected into the Jars MANIFEST file, cron-utils is sticked to Guava 18.X in an OSGi environment, although it could use any later release. At least up to 23.4-jre.

Since I'm not sure how to declare the range in maven I'm not going to add a PR. However I guess changing the POMs Guava version from

<guava.version>18.0</guava.version>

to

<guava.version>[18.0,)</guava.version>

could do the job.

Hi! First of all, thanks for your great work! I need help for maybe an issue when getting last execution of a UNIX cron expression (SingleExecutionTime implementation).

Look at this example: ExecutionTime.forCron(new CronParser(CronDefinitionBuilder.instanceDefinitionFor(UNIX)).parse("0 0 * * WED")).lastExecution(OffsetDateTime.of(2019, 6, 12, 0, 0, 0, 123, UTC).toZonedDateTime)

I think there is an issue when dealing with milliseconds. I was expecting that, in the given example, the lastExecution date is "2019-06-12T00:00:00.000Z", but actually is "2019-06-05T00:00:00.000Z", a week before. If you notice, the given datetime to evaluate lastExecution is not exactly at 0 milliseconds, instead is a bit later :) (123 millis). So, I think that the correct result is the closest previous execution, at 0 milliseconds exactly.

Look inside the code, maybe the problem is that potentialPreviousClosestMatch method is returning exactly the same inpute datetime when the former one matches the given cron expression. What do you think if, knowing that the calculation does not require to consider milliseconds, then the last line: return new ExecutionTimeResult(date, true) truncates milliseconds to 0.

I think this is a more correctly answer when dealing with lastExecution (maybe the problem is also when fetching nextExecution).

Regards, Patricio.

I use cronutils version 3.6.1. The Crontype.QUARTZ pattern "0 0/30 * * * ?" fails on

CronDefinition cronDefinition = CronDefinitionBuilder.instanceDefinitionFor( CronType.QUARTZ );
CronParser parser = new CronParser( cronDefinition );
Cron cron = parser.parse( "0 0/30 * * * ?" );
ExecutionTime time = ExecutionTime.forCron( cron );
Duration timeToNextExecution = time.timeToNextExecution( DateTime.now() );

with exception

14:22:35,227 ERROR [STDERR] com.cronutils.model.time.generator.NoSuchValueException
14:22:35,228 ERROR [STDERR]     at com.cronutils.model.time.generator.EveryFieldValueGenerator.generateNextValue(EveryFieldValueGenerator.java:43)
14:22:35,228 ERROR [STDERR]     at com.cronutils.model.time.generator.EveryFieldValueGenerator.generateCandidatesNotIncludingIntervalExtremes(EveryFieldValueGenerator.java:67)
14:22:35,229 ERROR [STDERR]     at com.cronutils.model.time.generator.FieldValueGenerator.generateCandidates(FieldValueGenerator.java:55)
14:22:35,229 ERROR [STDERR]     at com.cronutils.model.time.ExecutionTimeBuilder.forMinutesMatching(ExecutionTimeBuilder.java:51)
14:22:35,230 ERROR [STDERR]     at com.cronutils.model.time.ExecutionTime.forCron(ExecutionTime.java:92)
14:22:35,230 ERROR [STDERR]     at de.psi.messaging.purge.MessagePurgeTrigger.getDelay(MessagePurgeTrigger.java:35)

Reproduced by:

  @Test
  public void infiniteLoop() {
    Cron cron = new CronParser(CronDefinitionBuilder.instanceDefinitionFor(CronType.UNIX))
      .parse("0 0 30 2 *");
    ExecutionTime.forCron(cron).nextExecution(ZonedDateTime.now());
  }

I guess this should not happen?

https://github.com/jmrozanec/cron-utils/blob/dfbf4127afd2c28b6e44a76e2379be59e6137556/src/main/java/com/cronutils/model/time/generator/EveryFieldValueGenerator.java#L70

com.cronutils.model.time.generator.NoSuchValueException
    at com.cronutils.model.time.generator.EveryFieldValueGenerator.generateNextValue(EveryFieldValueGenerator.java:43)
    at com.cronutils.model.time.generator.EveryFieldValueGenerator.generateCandidatesNotIncludingIntervalExtremes(EveryFieldValueGenerator.java:67)
    at com.cronutils.model.time.generator.FieldValueGenerator.generateCandidates(FieldValueGenerator.java:55)
    at com.cronutils.model.time.ExecutionTimeBuilder.forMinutesMatching(ExecutionTimeBuilder.java:51)
    at com.cronutils.model.time.ExecutionTime.forCron(ExecutionTime.java:92)

Current implementation was developed using joda-time, since provides great date and time utils. A new date, time and calendar API was released at Java 8. Since we want to keep compatibility with previous JDK versions, we want to maintain a branch with joda-time dependency, and another to leverage the new Java 8 API.

cron-utils 9.0.1

reproducer (scala)

  val cronDefinition = CronDefinitionBuilder.instanceDefinitionFor(CronType.QUARTZ)
  val parser = new CronParser(cronDefinition)
  val cron = parser.parse("0 0 0 1 1 ? 2000")
  ExecutionTime.forCron(cron)

output

We for an class com.cronutils.model.time.generator.OnFieldValueGenerator builder

A cron expression for every 15th and last day of the month at 12:00 does not behave as expected and skips the last days when asked for next execution time.

Example cron schedule: 0 0 L,15 * *

While the cron expression is "valid" I wonder if this is in fact not supported.

Hi,

First of all thanks very much for this library. Only recently discovered it but it is extremely useful for us while moving away from a certain scheduler and replacing it with another one.

Our expressions typically contain written weekdays such as MON, TUE, etc. Parsing that up works great, but I can not find a way to get the output as such when building a CRON and getting the expression. (Cron#asString) It always seems to output integer values for the weekdays, which unfortunately, we can't deal with in our legacy solution.

Any idea / hints on how to get named weekdays instead of integers?

Thanks very much!

Some cron parsers do support intervals like @every 1h30m10s

For example: https://pkg.go.dev/github.com/robfig/cron#hdr-Intervals https://github.com/robfig/cron/blob/master/doc.go#L26

Would be great to get them in cron-utils too

Cron expressions like -> (*/2 * * * *) are supposed to run every two minutes from a scheduler that checks every 10 second if now is the execution time using

ExecutionTime.isMatch(ZonedDateTime zd)

Issue : In the 2nd , 4th , 6th ...... minute, the isMatch(date) method returns true for each of the 10th second in that particular minute. So cron expression runs 6 time instead of 1 time.

So my query was, is it like that only ? or I am making any mistake ?

Need help with this.

Note : Also I tried to build the code from latest branch(9.1.5) as well from the master branch. It didn't build properly. Tests are failing.

Given the following expressions:

cron1 = "0 0 10 1/1 ? "; //everyday at 10am
cron2 = "0 0 10 ? MON "; //every monday at 10am

We would ideally want to test cron1.overlap(cron2) -> true