[ISSUE-789] Add string functions TRIM, LPAD, RPAD#811
Conversation
Add three scalar string functions to GeaFlow DSL: - trim(str): strip leading/trailing whitespace - lpad(str, len, pad): left-pad string to target length - rpad(str, len, pad): right-pad string to target length Implementation: - Trim.java: extends UDF, supports both String and BinaryString - LPad.java: extends UDF, handles null/empty/truncate edge cases - RPad.java: extends UDF, handles null/empty/truncate edge cases - Registered all three in BuildInSqlFunctionTable.java - Added test SQL + expected output files for each function - Added StringFunctionTest test class Closes apache#789
| if (len <= 0) { | ||
| return ""; | ||
| } | ||
| if (len <= str.length()) { |
There was a problem hiding this comment.
String.length(), substring(), and StringBuilder.length() operate on UTF-16 code units rather than Unicode code points. This makes LPAD incorrect for supplementary characters and can split a surrogate pair while truncating.
For example, lpad("😀", 2, "x") should return "x😀" under MySQL's character-based semantics, but this implementation returns "😀".
| if (len <= 0) { | ||
| return ""; | ||
| } | ||
| if (len <= str.length()) { |
There was a problem hiding this comment.
This has the same UTF-16 code-unit issue as LPAD.
| if (s == null) { | ||
| return null; | ||
| } | ||
| return StringUtils.strip(s); |
There was a problem hiding this comment.
StringUtils.strip(s) removes tabs and other Java whitespace characters, while the BinaryString overload below removes only the ASCII space character. This creates overload-dependent behavior.
MySQL's default TRIM behavior removes spaces, so trim("\thello\t") should retain the tabs, but this overload currently returns "hello".
|
|
||
| INSERT INTO output_console | ||
| SELECT | ||
| lpad('hi', 5, 'x'), |
There was a problem hiding this comment.
Could we extend the runtime tests with supplementary Unicode characters, null arguments, an empty padding string, and zero/negative target lengths?
|
|
||
| INSERT INTO output_console | ||
| SELECT | ||
| trim(' hello '), |
There was a problem hiding this comment.
Please add a case containing tabs, such as trim('\thello\t'), to define and verify the intended semantics. It would also be useful to cover null input here.
2 participants
What does this PR do?
Closes #789
Adds three string scalar functions to GeaFlow DSL:
trim(str)— strip leading/trailing whitespacelpad(str, len, pad)— left-pad to target lengthrpad(str, len, pad)— right-pad to target lengthKey design decisions
Verification
SELECT trim(' hello '); -- hello
SELECT lpad('hi', 5, 'x'); -- xxxhi
SELECT rpad('hi', 5, 'x'); -- hixxx