jin-won-park opened a new issue, #18922:
URL: https://github.com/apache/iceberg/issues/18922

   ### Apache Iceberg version
   
   main (development)
   
   ### Query engine
   
   None
   
   ### Please describe the bug 🐞
   
   ### Problem
   
   `truncate[W]` on strings truncates to W code points 
(`UnicodeUtil.truncateString`), but `TruncateString.project` and 
`projectStrict` decide how to project `startsWith`/`notStartsWith` by comparing 
the prefix's `String.length()` (UTF-16 chars) with W.
   
   A prefix with supplementary characters such as emoji can have `length() == 
W` and still have fewer than W code points. The projection then treats it as a 
full partition value and turns it into an equality, while the partition values 
of matching rows are longer.
   
   ### Reproduce
   
   ```java
   Schema schema = new Schema(Types.NestedField.optional(1, "s", 
Types.StringType.get()));
   PartitionSpec spec = PartitionSpec.builderFor(schema).truncate("s", 
2).build();
   
   // "😀" is 1 code point but 2 chars. A row "😀abc" lands in partition s_trunc 
= "😀a".
   Projections.inclusive(spec).project(Expressions.startsWith("s", "😀"));
   // actual:   ref(name="s_trunc") == "😀"         -> partition "😀a" is pruned
   // expected: ref(name="s_trunc") startsWith "😀"
   
   Projections.strict(spec).project(Expressions.notStartsWith("s", "😀"));
   // actual:   ref(name="s_trunc") != "😀"         -> partition "😀a" is treated 
as all-matching
   // expected: ref(name="s_trunc") notStartsWith "😀"
   ```
   
   ### Impact
   
   - Inclusive projection prunes partitions that contain matching rows, so 
scans with such a `startsWith` filter silently return fewer rows.
   - Strict projection reports partitions as fully matching when none of their 
rows match, which affects residuals and partition-level decisions such as 
metadata-only deletes.
   
   It only happens for string `truncate` partitions when the prefix's UTF-16 
length equals the width but its code point count is smaller. ASCII and BMP-only 
prefixes are unaffected.
   
   ### Proposed fix
   
   Compare the prefix's code point count with the width in both methods. I have 
a fix with tests and can open a PR.
   
   ### Willingness to contribute
   
   - [x] I can contribute a fix for this bug independently
   - [ ] I would be willing to contribute a fix for this bug with guidance from 
the Iceberg community
   - [ ] I cannot contribute a fix for this bug at this time


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to