Java error from getDbCovariateData

Hello!

After many months of work without this bug, we've recently come across a bug on two separate computers that seem to originate from java.

"Error in .jcall("RJavaTools", "Ljava/lang/Object;", "invokeMethod", cl, :
java.lang.ClassCastException: class java.lang.String cannot be cast to class java.lang.Integer (java.lang.String and java.lang.Integer are in module java.base of loader 'bootstrap')"

We feed a cohort table to getDbCohortMethodData, which calls getDbCovariateData, which in turn calls getDbDefaultCovariateData. Running the code there line-by-line points toward the following call:

rJava::J("org.ohdsi.featureExtraction.FeatureExtraction")$createSql(settings, aggregated, cohortTable, rowIdField, java_array, cdmDatabaseSchema)

We're running sql server, this is the java-version information:

• openjdk version "16.0.1" 2021-04-20
• OpenJDK Runtime Environment AdoptOpenJDK-16.0.1+9 (build 16.0.1+9)
• OpenJDK 64-Bit Server VM AdoptOpenJDK-16.0.1+9 (build 16.0.1+9, mixed mode, sharing)

The input arguments to this rJava::J-function are not very helpful as reproducible example, since a restart of R and feeding the same argument gives you a different error message ("Error in jclassName(class, class.loader = class.loader) : java.lang.ClassNotFoundException"). But here are our input arguments:

settings = '{\"temporal\":false,\"DemographicsGender\":true,\"DemographicsAge\":true,\"ConditionOccurrenceLongTerm\":true,\"ConditionGroupEraLongTerm\":true,\"DrugGroupEraLongTerm\":true,\"longTermStartDays\":-180,\"mediumTermStartDays\":-180,\"shortTermStartDays\":-90,\"endDays\":0,\"includedCovariateConceptIds\":[],\"addDescendantsToInclude\":true,\"excludedCovariateConceptIds\":[\"31317\",\"1139699\"],\"addDescendantsToExclude\":true,\"includedCovariateIds\":[]}' aggregated = FALSE cohortTable = "#cohort_person" rowIdField = "subject_id" java_array = rJava::.jarray(as.character(-1)) cdmDatabaseSchema = "OmopCdm.synpuf5pct_20180710"

If we instead feed a drug_era and condition_era instead of cohort-tables to getDbCohortMethodData, things are working again.
I've also tried to reproduce the bug starting from Eunomia, but the bug does not appear there.

`connection <- DatabaseConnector::connect(Eunomia::getEunomiaConnectionDetails())
DatabaseConnector::executeSql(connection, "DROP TABLE IF EXISTS cohort_table")
DatabaseConnector::executeSql(connection, "CREATE TABLE cohort_table (SUBJECT_ID INT, COHORT_DEFINITION_ID INT, COHORT_START_DATE DATETIME, COHORT_END_DATE DATETIME)")
DatabaseConnector::executeSql(connection, "INSERT INTO cohort_table(SUBJECT_ID, COHORT_DEFINITION_ID, COHORT_START_DATE, COHORT_END_DATE) VALUES (1, 1, '2000-01-01 12:00:00', '2000-01-02 12:00:00')")
DatabaseConnector::querySql(connection, "SELECT * FROM cohort_table")
covSettings <- FeatureExtraction::createCovariateSettings(useConditionGroupEraLongTerm = TRUE, excludedCovariateConceptIds = c(31317, 1139699))

covariateData <- FeatureExtraction::getDbCovariateData(connection = connection,
cdmDatabaseSchema = "main",
cdmVersion = 5,
cohortTable = "cohort_table",
cohortTableIsTemp = FALSE,
rowIdField = "SUBJECT_ID",
covariateSettings = covSettings)
`

Any ideas or input on how to solve this would be highly appreciated. Have there been any recent updates to FeatureExtraction/Java-related matters the last week?

Start at getDbCohortMethodData, getDbCovariateData, getDbDefaultCovariateData, and the rJava FeatureExtraction$createSql call, using the supplied SQL Server arguments and Java 16 details. Compare the cohort-table path with drug_era/condition_era and the Eunomia reproduction; done means the reported input no longer produces the Java ClassCastException or ClassNotFoundException.