fix(lib-storage): keep empty fields when splitting per-line records - #1521

Summary

The per-line transportOneRecord(RecordSender, Configuration, TaskPluginCollector, String) overload (the hdfsreader sequence-file path, invoked once per record from DFSUtil.sequenceFileStartRead) split record lines with StringUtils.split, which drops empty tokens and trailing empties.

A Hive/sequence-file row with an empty column (e.g. a,,b) was split to [a, b], shifting every subsequent column against the configured index/type mapping — values landed in the wrong typed column or the row was thrown away as dirty with an IndexOutOfBoundsException. The same content read through the commons-csv path (doReadFromStream, fileType=text) keeps the empty fields, so behavior was inconsistent by file type and data was silently corrupted on the seq path.

What type of change is this?

• Bugfix

Changes

• Replaced StringUtils.split(line, fieldDelimiter) with a small indexOf-based splitter that keeps every token, including empty and trailing-empty ones — the same semantics commons-csv produces on the doReadFromStream path.

Motivation and Context

Reported against lib/addax-storage by a module code review (cross-file contract tracer verified the divergence between the seq per-line path and the commons-csv path).

How has this been tested?

• mvn -DskipTests compile -pl :addax-storage (JDK 17) — clean.
• Manual trace: "a,,b" → 3 tokens, "a,b," → 3 tokens, no-delimiter line → 1 token.
• No unit tests added; per project convention, verification happens in the build environment manually.

Checklist (required)

• I have read the project's contributing guidelines and code of conduct.
• My change includes appropriate tests (if applicable). — n/a, manual verification per project convention
• I ran a local build and fixed any compilation issues.
• I updated or added documentation if needed (README, docs/ or docs/en/). — n/a
• I updated CHANGELOG.md when appropriate.
• I have not added any secrets or credentials.
• I have checked for sensitive or personal data in this PR.