Known Limitations

This document applies to Materialize v0.9.5.

The following links describe features that Materialize does not yet support, but plans to in future releases. If any of these issues impact you, feel free to let us know on the linked-to GitHub issue.

For a complete and current overview, check out our known-limitations tag.

SQL

Types

• Numeric precision is not always equivalent to PostgreSQL (#1824)
• Numeric to float conversions are susceptible to floating point errors (#1102)

Performance

• ANY and ALL queries generate suboptimal plans (#744)

Functions

• Materialize does not support window functions. In some cases, you may be able to achieve the desired results with temporal filters or the TOP K by group idiom instead. (#213)

Common table expressions (CTEs)

• CTEs only support SELECT queries. (#4867)
• Materialize inlines the CTE where it’s referenced, which could cause unexpected performance characteristics for especially complex expressions. (#4867)
• WITH RECURSIVE CTEs are not available yet. (#2516)

Sources & sinks

Kafka

• Protobuf data for Kafka sinks is not supported (#1541)

Kinesis

• Consistent Kinesis sources are not supported (#2191)
• Enhanced fan-out Kinesis consumers are not supported (#2192)
• Reading from a Kinesis stream as of a sequence number or timestamp is not supported (#2294)
• Kinesis sinks are not supported (#2372)

File sources

None right now!

Regex formatting

• Regex columns cannot be set as nullable (#1685)

Deployment

Performance

• Slow queries can impact other, unrelated queries (#1956)

Networking

• Hostnames that resolve to multiple addresses are not supported (#502)

Monitoring & debugging

• No report exists to show memory usage for indexes (#1532)

Closed known limitations

The following issued used to be known limitations, but we’ve fixed them in the specified version.