rdbc-io / rdbc Goto Github PK

Publisher verification tests should work on mocks, not on the real connection.

Common for all statements

Errors

• non-existent table
• non-existent column
• missing param value
• superfluous param
• syntactically incorrect
• timeout
• positional params ^
• can't mix positional with named params
• TODO don't forget to test all the cases above for executing batch

Selects - everything inside a tx and without an explicit tx

Streaming

• Select empty table
• Read fully at once
• Read fully by chunks
• Read some & cancel
• Cancel right away
• warnings
• rows affected count
• metadata

Result set

• Select empty table
• Select non-empty table
• warnings
• rows affected count
• metadata

For first row

• Empty table
• Single row
• Many rows

For value

• Empty table
• Single row
• Many rows
• Incorrect type
• null value

For value opt

• Empty table
• Single row
• Many rows
• Incorrect type
• null value

Streaming arguments

• TODO
• Don't forget about arg conversion errors and execution errors in the middle of streaming
• Test what happens when publisher fails: connection should roll back

Row

• opt/non opt

Data types

outbound

• ...

inbound

• ...

Param tests

Do select with different parameters and see if results show that parameters got applied

• no parameters
• one parameter
• many parameters
• the same parameter multiple times
• parameter at the end of sql

TX tests

...

Cross tests with JDBC

...

Decide whether ResultStream containing metadata & publisher should be returned from method "executeForStream" or Publisher directly. Maybe something that extends a Publisher containing all the metadata could be returned?

The work will be performed in a dedicated repository: https://github.com/rdbc-io/rdbc-pool

Maybe support positional parameters to make it easier to implement?

Currently tests are organized by feature. Consider organizing them by class.

The current tests cover the code but don't check for the functionality of tracing - fix this.

Introduce "whenIdle" method on connection that will start executing stuff after the conn becomes ready and the code being executed will "own" the connection so that no other request can interfere.

1. Figure out how to use normal and transactional loan pattern is possible with streaming.
2. Introduce maybe some flag for stream method: closeConn or something that will close collection on stream cancellation, error or completion.

Describe the solution in the docs.

• LocalDate
• Boolean
• Char
• BigDecimal
• Double
• Float
• Instant
• LocalDateTime
• ZonedDateTime
• LocalTime
• Array[Byte]
• byte stream?

Decide whether Java API should return CompletionStage<? extends Connection> etc. instead of CompletionStage<Connection>.