fix of overflow detection for numeric primitive types #121
Conversation
Codecov Report
@@ Coverage Diff @@ ## master #121 +/- ## ========================================== - Coverage 68.53% 68.38% -0.15% ========================================== Files 107 107 Lines 7084 7101 +17 Branches 1347 1376 +29 ========================================== + Hits 4855 4856 +1 - Misses 1800 1806 +6 - Partials 429 439 +10
Continue to review full report at Codecov.
|
| if (value == Long.MIN_VALUE) { | ||
| throw iter.reportError("readLongSlowPath", "value is too large for long"); | ||
| if (value < multmin) { | ||
| throw iter.reportError("readIntSlowPath", "value is too large for int"); |
There was a problem hiding this comment.
throw iter.reportError("readIntSlowPath", "value is too large for int");
should be long
There was a problem hiding this comment.
fixed by following 2 commits
please sqash them in case of merge
| value = (value << 3) + (value << 1) + ind; | ||
| if (value < 0 && value != Long.MIN_VALUE) { | ||
| value = (value << 3) + (value << 1); | ||
| if (value < limit + ind) { |
There was a problem hiding this comment.
may be we can change
if (value < limit + ind) { throw iter.reportError("readLongSlowPath", "value is too large for long"); } value -= ind;to
value -= ind; if (value >= 0) { throw iter.reportError("readLongSlowPath", "value is too large for long"); }if value is positive, we add additional value != Long.MIN_VALUE check
There was a problem hiding this comment.
limit + ind is required to properly detect overflow when number to parse is Long.MAX_VALUE + 1 to do not raise overflow when parsing Long.MIN_VALUE
| } | ||
| | ||
| static long readLongSlowPath(JsonIterator iter, long value) throws IOException { | ||
| static long readLongSlowPath(final JsonIterator iter, long value, final boolean negative) throws IOException { |
There was a problem hiding this comment.
final boolean negative is not necessary. we can always return negative number, and make it positive from the readLong. this way saved one branch.
There was a problem hiding this comment.
some additional param is required until we inline these slow path methods or split them on pair of methods (one for negative, other for positive)
Problem is that 40000000000000000000 (too big to be mapped to java.langLong number in JSON) is succesfully parsed to 3106511852580896768L using current implementation from develop.