Merge pull request ClickHouse#87258 from ClickHouse/pq0

al13n321 · mkmkme · commit 8bde797fe8d5 · 2026-02-09T13:37:41.000+01:00
Fix 'Invalid dict indices bit width: 0' in parquet reader v3
diff --git a/src/Processors/Formats/Impl/Parquet/Decoding.cpp b/src/Processors/Formats/Impl/Parquet/Decoding.cpp
@@ -41,7 +41,7 @@ struct BitPackedRLEDecoder : public PageDecoder
             requireRemainingBytes(1);
             bit_width = size_t(UInt8(*data));
             data += 1;
-            if (bit_width > 8 * sizeof(T) || (bit_width == 0 && limit > 1))
+            if (bit_width > 8 * sizeof(T))
                 throw Exception(ErrorCodes::INCORRECT_DATA, "Invalid dict indices bit width: {}", bit_width);
         }
         else
@@ -120,7 +120,7 @@ struct BitPackedRLEDecoder : public PageDecoder
     {
         if (bit_width == 0)
         {
-            /// bit_width == 0 can be used for dictionary indices if the dictionary has only one value.
+            /// bit_width == 0 means all values are 0.
             if constexpr (!skip)
                 memset(out, 0, num_values * sizeof(T));
             return;

Original file line number	Diff line number	Diff line change
`@@ -41,7 +41,7 @@ struct BitPackedRLEDecoder : public PageDecoder`
`41`	`41`	`requireRemainingBytes(1);`
`42`	`42`	`bit_width = size_t(UInt8(*data));`
`43`	`43`	`data += 1;`
`44`		`- if (bit_width > 8 * sizeof(T) \|\| (bit_width == 0 && limit > 1))`
	`44`	`+ if (bit_width > 8 * sizeof(T))`
`45`	`45`	`throw Exception(ErrorCodes::INCORRECT_DATA, "Invalid dict indices bit width: {}", bit_width);`
`46`	`46`	`}`
`47`	`47`	`else`
`@@ -120,7 +120,7 @@ struct BitPackedRLEDecoder : public PageDecoder`
`120`	`120`	`{`
`121`	`121`	`if (bit_width == 0)`
`122`	`122`	`{`
`123`		`- /// bit_width == 0 can be used for dictionary indices if the dictionary has only one value.`
	`123`	`+ /// bit_width == 0 means all values are 0.`
`124`	`124`	`if constexpr (!skip)`
`125`	`125`	`memset(out, 0, num_values * sizeof(T));`
`126`	`126`	`return;`