E5 - Reporter API & assertion output improvements

[jgebal]

Target release or timeframe

No response

Scope and goals

Problems

Coverage reporters, especially the HTML coverage reporter, produce a single output file.
That approach doesn't scale to schemas with thousands of packages and causes may problems with memory consumption.

When running expectations on cursor and collection data, the comparison is more complex than needed. For a cursor with scalar columns a plain check for non-intersecting elements is sufficient. This will significantly improve the performance of data comparison.
The diff data is returned diffs in XML while it could be returned as JSON.

The reporter output is now capped at returning 4k chars per one line which affects readability of comparison.
Most commonly affected are the cursor, object, collection as well as clob and varchar comparison.

Proposed direction

• Rework the reporter API so consumers can split output across multiple files.
• Optimize cursor comparison for simple cursor shapes.
• switch cursor diff output to JSON.
• remove the 4K limit for the expectation result output size.

Breaking change

Could be breaking change for reporter API consumers unless the new output is provided in another format.
The change of diff format change (XML → JSON) will affect anything parsing the current diff output.

Tracking

• Reporter API: support splitting output across multiple files utPLSQL#1372
• Cursor and collection comparison: performance for simple data utPLSQL#1373
• Cursor and object comparison: JSON output utPLSQL#1374
• Expectation result: remove 4K truncation utPLSQL#1375

Open questions for the community

No response