Releases of Oracle R Enterprise often contain new features. The features in the current release and in some previous releases are described in the following topics:
Oracle R Enterprise 1.5 introduces functions for managing Oracle R Enterprise datastores and scripts in the Oracle Database script repository. It also contains a new function in the OREmodel package, new transparency layer methods for some functions in the stats package, and other enhancements.
Oracle R Enterprise 1.5 has new R functions and SQL procedures, new transparency layer methods, and enhancements to some functions.
Oracle R Enterprise 1.5 has new R functions and SQL procedures for managing Oracle R Enterprise datastores and the Oracle R Enterprise R script repository. Users can now share with other users access to datastores and registered R scripts. This release also has a new modeling function, ore.randomForest, new svd and prcomp statistical function methods that take ore.frame objects and can use parallel processing, and other enhancements.
The following topics briefly describe the new features.
Oracle R Enterprise 1.5 provides new R functions for managing Oracle R Enterprise datastores and the R script repository.
The owner of a datastore or registered R script can now share with other users read privilege access to the datastore or script. This release also has new arguments to functions that create datastores and scripts and that give information about them. This datastore and script management functionality has both R and SQL interfaces.
The R functions for managing datastores are the following:
ore.delete, which deletes a datastore, is unchanged.
ore.grant is a new function that grants read privilege access to a datastore or script.
ore.load, which loads objects from a datastore into an R environment, has the new argument owner that specifies the datastore owner.
ore.revoke is a new function that revokes read privilege access to a datastore or script.
ore.save, which creates a datastore, has the new argument grantable that specifies whether read access can be granted to the datastore.
ore.datastore, which lists information about a datastore, has the new argument type that specifies the type of datastore. The values of type are the character strings user (the default), grant, granted, and all.
ore.datastoreSummary, which provides detailed information about datastores, has the new argument owner that specifies a datastore owner.
The R functions for managing scripts are the following:
ore.grant is a new function that grants read privilege access to a datastore or script.
ore.revoke is a new function that revokes read privilege access to a datastore or script.
ore.scriptCreate, which adds an R function to the script repository, has the new arguments global and overwrite. The argument global specifies whether the script is global or private. A global script is a public script that is available to all users. If global = FALSE, then access to the script must be granted by the owner to other users. The argument overwrite specifies whether the content of a script can be replaced.
ore.scriptDrop, which deletes a script, has the new arguments global, which specifies whether the script to drop is global or not, and silent, which specifies whether to report an error if the script cannot be dropped.
ore.scriptList, which lists information about scripts, has the new argument type that specifies the type of script. The values of type are the character strings user (the default), global, grant, granted, and all.
ore.scriptLoad is a new function that loads a script into the R environment.
Oracle R Enterprise 1.5 provides new PL/SQL procedures for managing Oracle R Enterprise datastores and the R script repository.
The owner of a datastore or registered R script can now share with other users read privilege access to the datastore or script. This release also has new arguments to procedures that create datastores and scripts and that give information about them. This functionality has both R and SQL interfaces. Oracle Database data dictionary views provide information about datastores and scripts.
The SQL procedures for controlling access to Oracle R Enterprise datastores and registered R scripts are described in the following sections:
PL/SQL Procedures for Managing Datastores
The PL/SQL procedures for managing datastores are the following:
rqDropDataStore, which deletes a datastore, is unchanged.
rqGrant is a new procedure that grants read privilege access to a datastore or script.
rqRevoke is a new procedure that revokes read privilege access to a datastore or script.
PL/SQL Procedures for Managing Scripts
The PL/SQL procedures for managing scripts are the following:
rqGrant is a new procedure that grants read privilege access to a datastore or script.
rqRevoke is a new procedure that revokes read privilege access to a datastore or script.
rqScriptCreate has the new arguments global and overwrite. The argument global specifies whether the script is global or private. A global script is a public script that is available to all users. If global = FALSE, then access to the script must be granted by the owner to other users. The argument overwrite specifies whether the content of a script can be replaced.
rqScriptDrop has the new arguments global, which specifies whether the script to drop is global or not, and silent, which specifies whether to report an error if a script cannot be dropped.
Data Dictionary Views for Datastores
The Oracle Database dictionary views related to datastores are the following:
ALL_RQ_DATASTORES
RQUSER_DATASTORECONTENTS
USER_RQ_DATASTORES
USER_RQ_DATASTORE_PRIVS
Data Dictionary Views for Scripts
The Oracle Database dictionary views related to scripts are the following:
ALL_RQ_SCRIPTS
USER_RQ_SCRIPTS
USER_RQ_SCRIPT_PRIVS
Function ore.groupApply now supports partitioning on multiple columns.
The INDEX argument can now take an ore.vector or ore.frame object that contains ore.factor objects or columns, each of which is the same length as argument X. Function ore.groupApply uses the INDEX object to partition the data in X before sending it to function FUN.
For an example of the use of the INDEX object to partition the data, see “Partitioning on Multiple Columns”.
The ore.randomforest function builds a random forest model on data in an ore.frame object.
It uses embedded R execution to grow random forest trees in parallel in R sessions on the database server. It returns an ore.randomforest object. In Oracle R Enterprise 1.5, function ore.randomForest supports classification but not regression.
The ore.randomforest function uses the same algorithm as that adopted by the CRAN R randomForest package but it has better runtime memory usage as well as ensemble tree size.
The scoring method predict on an ore.randomforest model also runs in parallel. Oracle recommends that you set the cache.model argument to TRUE when sufficient memory is available. Otherwise, you should set cache.model to FALSE to prevent memory overuse.
To use ore.randomforest, you must install either Oracle R Distribution (ORD) 3.2 or the CRAN R randomForest package. Oracle recommends that you use the function ore.randomforest in ORD 3.2, which offers better performance and scalability than the CRAN R randomForest. If you only install the R randomForest package, ore.randomForest issues a warning message at run time. The CRAN R randomForest package is one of the supporting packages in Oracle R Enterprise 1.5.
The global option ore.parallel determines the degree of parallelism to use in the Oracle R Enterprise server. The argument groups controls the granularity of the ore.randomForest model.
For an example of using ore.randomForest, see "Building a Random Forest Model."
Function ore.summary has improved performance. It also has a different signature and the data types for some arguments have changed.
The function’s syntax is now the following:
ore.summary(data, var, stats = c("n", "mean", "min", "max"),
class = NULL, types = NULL, ways = NULL, weight = NULL,
order = NULL, maxid = NULL, minid = NULL, mu = 0,
no.type = FALSE, no.freq = FALSE)
The differences between the ore.summary in Oracle R Enterprise Release 1.5 and previous releases are the following:
The performance of ore.summary has improved; it now returns over an order of magnitude faster.
The arguments var, stats, and class now take a vector of character strings; previously, they took a concatenated string using a comma as the separator.
Argument types is now a list of character string vectors that specifies combinations of columns in the class argument.
Arguments maxid and minid are named vectors of character strings.
Arguments group.by and no.level are not supported.
Argument mu in previous releases was named mu0.
For examples of using ore.summary, see “Summarizing Data with ore.summary”.
Oracle R Enterprise 1.5 provides transparency layer methods for the stats package functions prcomp and svd.
The prcomp function performs principal components analysis and the svd function performs singular-value decomposition. Those functions now accept ore.frame objects and can use parallel execution in the database, which can improve scalability and performance.
Some Oracle R Enterprise functions now support the Oracle Database data types BLOB and CLOB.
Functions ore.push and ore.pull now support the database data types BLOB and CLOB.
Embedded R execution R functions now support the database data types BLOB and CLOB for input and output objects.
For examples of using BLOB and CLOB data types, see "Example 6-11".
The following topics describe the changes in Oracle R Enterprise 1.4:
The following changes are in Oracle R Enterprise 1.4.1:
The ore.glm function now accepts offset terms in the model formula and the function can now be used to fit negative binomial and tweedie families of generalized linear models.
The ore.sync function has an additional optional argument, query, that creates an ore.frame object from an optimized SQL SELECT statement without creating a view in the database. You can use this argument to create a query even when you do not have the CREATE VIEW system privilege for the current schema.
See Also:
The new global option for serialization, ore.envAsEmptyenv, specifies whether referenced environments in an object should be replaced with an empty environment during serialization to an Oracle Database. This option is used by the following functions:
ore.push, which for a list object accepts envAsEmptyenv as an optional argument
ore.save, which has envAsEmptyenv as a named argument
ore.doEval and the other embedded R execution functions, which accept ore.envAsEmptyenv as a control argument.
The default values of the above arguments are regulated by the global option ore.envAsEmptyenv, but by using the argument you can override the global option value for a function.
See Also:
The online help for the ore.push and ore.save functions
The following topics describe the changes in Oracle R Enterprise 1.4:
The following changes are in Oracle R Enterprise 1.4:
Additions and improvements to data preparation functions:
The new factanal function performs factor analysis on a formula or an ore.frame object that contains numeric columns.
Both signatures of the princomp function support the scores, subset, and na.action arguments.
The new getXlevels function creates a list of factor levels that can be used in the xlev argument of a model.matrix call that involves an ore.frame object.
The new exploratory data analysis function ore.esm builds exponential smoothing models for time series data. The function builds a model using either the simple exponential smoothing method or the double exponential smoothing method. The function can preprocess the time series data with operations such as aggregation and the handling of missing values. See "Building Exponential Smoothing Models on Time Series Data".
Additions and improvements to the Oracle R Enterprise regression and neural network modeling functions:
The new ore.glm function provides methods for fitting generalized linear models, which include logistic regression, probit regression, and poisson regression. See "Building a Generalized Linear Model".
The ore.lm and ore.stepwise functions are no longer limited to a total of 1,000 columns when deriving columns in the model formula.
The ore.lm function now supports a weights argument for performing weighted least squares regression.
The anova function can now perform analysis of variance on an ore.lm object.
For the ore.stepwise function, the values for the direction argument have changed. The value "both" now prefers drops over adds. The new direction argument value "alternate" has the previous meaning of the "both" value.
The ore.neural function has several new arguments.
Additions and improvements to the Oracle Data Mining model algorithm functions:
The new ore.odmAssocRules function, which builds an Oracle Data Mining association model using the apriori algorithm. See "Building an Association Rules Model".
The new ore.odmNMF function, which builds an Oracle Data Mining model for feature extraction using the Non-Negative Matrix Factorization (NMF) algorithm. See "Building a Non-Negative Matrix Factorization Model".
The new ore.odmOC function, which builds an Oracle Data Mining model for clustering using the Orthogonal Partitioning Cluster (O-Cluster) algorithm. See "Building an Orthogonal Partitioning Cluster Model".
An additional global option for Oracle R Enterprise, ore.parallel. See "Oracle R Enterprise Global Options".
The following topics describe the changes in Oracle R Enterprise 1.3:
The new features in Oracle R Enterprise 1.3 are the following:
Predicting with R models using in-database data with the OREpredict package
Ordering and indexing with row.names<-
Predicting with Oracle Data Mining models using the OREodm package
Saving and managing R objects in the database
Date and time data types
Sampling and partitioning
Long names for columns
Automatically connecting to an Oracle Database instance in embedded R scripts
Building an R neural network using in-database data with the ore.neural function
Other changes in this release are the following:
Installation and administration information has moved from this manual to Oracle R Enterprise Installation and Administration Guide. New features related to installation and administration are described in that book.
The new features in Oracle R Enterprise 1.1 are the following:
Support for additional operating systems:
Oracle R Distribution and Oracle R Enterprise are now supported IBM AIX 5.3 and higher and on 10 and higher for both 64-bit SPARC and 64-bit x386 (Intel) processors.
The Oracle R Enterprise Server now runs on 64-bit and 32-bit Windows operating systems.
Improved mathematics libraries in R:
You can now use the improved Oracle R Distribution with support for dynamically picking up either the Intel Math Kernel Library (MKL) or the AMD Core Math Library (ACML) with Oracle R Enterprise.
On Solaris, Oracle R Distribution dynamically links with Oracle SUN performance library for high speed BLAS and LAPACK operations.
Support for Oracle Wallet enables R scripts to no longer need to have database authentication credentials in clear text. Oracle R Enterprise is integrated with Oracle Wallet for that purpose.
Improved installation scripts provide more prerequisite checks and detailed error messages. Error messages provide specific instructions on remedial actions.