New in InterSystems IRIS 2025.1

This release of InterSystems IRIS introduces the capability to leverage Python statements in your Data Transformation Language (DTL) classes. For each action, users can select the language (ObjectScript or Python) enabling developers who are skilled in Python to utilize larger parts of the platform with their existing skill set.

InterSystems IRIS Vector Search now features a disk-based Approximate Nearest Neighbor (ANN) index for VECTOR—typed columns. The ANN index is based on the popular Hierarchical Networked Small Worlds (HNSW) algorithm. It yields sub-second similarity query response times over millions of vectors, providing a 2,500 times speedup in performance compared to the "brute force" non-indexed vector search, when searching over 100 million vectors.

When creating the ANN index, the user must choose which similarity function will be used (either VECTOR_DOT_PRODUCT or VECTOR_COSINE) to query the data. The ANN also supports incremental updates, and can be used in combination with the %EMBEDDING index.

This release includes changes to the InterSystems IRIS Business Intelligence cube building process to provide safeguards for updating related cube dependency chains and other enhancements that address several customer-reported issues in this area. To ensure that analysis of cube builds is always taken into consideration and that data remains consistent, the base functions for the DeepSee.CubeManager APIs (%BuildCube and %SynchronizeCube) now perform cube dependency analysis automatically and execute a full model group update. This new procedure includes:

• Analyzing the requested cube to see if there are any relationship dependencies and then internally establishing a full list of cubes that need to be updated, along with appropriate build orders for that cube list.

• Executing build and synchronization of individual cubes in succession for the list, as needed, using the new %BuildOneCube and %SynchronizeOneCube functions.

• Collating the singular statistics into a complete set to return to the caller.

These changes also include logging of all events recorded for all cubes built in the dependency chain, independent of the Cube Manager, making troubleshooting more accessible.

InterSystems IRIS SQL has long supported the TOP keyword to limit query results only to the first n records of a sorted result set. With this release, InterSystems adds support for two alternative syntax flavors: LIMIT ... OFFSET ..., which is commonly used in other database platforms, and OFFSET ... FETCH ..., which is the official ANSI standard. By offering all three methods of server-side pagination, InterSystems has made it easier for new customers and users migrating to InterSystems IRIS to write efficient queries.

Based on customer feedback on the LOAD DATA command, InterSystems IRIS SQL now supports a similar LOAD SQL command to load and execute any number of SQL DDL statements from a file or directory. This new command offers a simple entry point for developers accessing the system over SQL and writes detailed output, either to the SQL DiagnosticsOpens in a new tab table or another log file.

The pre-existing ObjectScript entry pointsOpens in a new tab have also been enhanced to support the diagnostic output.

This release extends the ALTER TABLE ... CONVERT command to support conversion between row and columnar table layouts. Users can choose to convert the whole table or a subset of columns. Queries and DML (INSERT, UPDATE, and DELETE) commands can still be executed while the table conversion is taking place. Please see ALTER TABLE for more details on the conversion process.

This release introduces two new convenience functions to the InterSystems SQL dialect:

• APPROX_COUNT_DISTINCT() is a new aggregate function that implements the HyperLogLog algorithm, which offers users an estimate of the number of distinct values for a particular column that is orders of magnitude faster than the exact result obtained from a COUNT(DISTINCT) command.

• HASHBYTES() offers a convenient SQL entry point for common hash functions, such as SHA256.

The 2025.1 release improves the scalability of $INCREMENT and introduces a new, more compact journal file format. The new format uses less I/O bandwidth (due to the smaller journal record size) on average and includes changes to the way that $INCREMENT is recorded in the journal. Journal files created on a 2025.1 instance cannot be opened on an instance running a prior release, so be sure to keep track of which mirror members are upgraded. Applications that have implemented custom logic to read and interpret journal files should be reviewed to determine whether this logic is impacted by the new $INCREMENT handling discussed in Special Considerations When Upgrading; most such custom logic is not impacted.

This release includes a number of enhancements to increate the efficiency of database compactionOpens in a new tab and defragmentation. Especially for large databases with lots of big strings, the compaction process now runs significantly faster and is better aligned with any concurrent block allocation for application workloads.

Previously, adding a database to a mirror required following a manual process copying the database file to the other mirror members, coordinated with API calls to register the database. Starting with this release, the new database can be created automatically on all mirror members and global data is downloaded automatically, without requiring file copies. These improvements significantly simplify the setup of complex mirrored environments, although attention should be paid to the size of the database being added to the mirror, as file copies may still be a practical solution when the database being added already contains a large amount of data. Please refer to the API reference for more details on the new option.

This release introduces a new ^ECP command line utility to perform common ECP management tasks, similar to existing ^SECURITY and ^DATABASE utilities. This means administrators of ECP environments no longer need to rely on the Management Portal for most configuration tasks.

InterSystems can now interact with the cryptographic libraries of FIPS 140–3. Previous versions supported FIPS 140–2, so this is an important upgrade for users that require the latest FIPS-compliant libraries.