fix(otlp): coerce empty and "-" trace attribute strings to null (#9461)

* fix(otlp): coerce empty and "-" trace attribute strings to null

Instrumentations fill numeric attributes from absent headers with an empty
string (PSR-7 getHeaderLine returns ""). Coercing that into an existing
numeric column failed and rejected the whole span.

Signed-off-by: Dennis Zhuang <killme2008@gmail.com>

* test: trim redundant row from empty-string rewrite test

Signed-off-by: Dennis Zhuang <killme2008@gmail.com>

---------

Signed-off-by: Dennis Zhuang <killme2008@gmail.com>

src/frontend/src/instance/otlp/trace_types.rs

@@ -519,31 +519,34 @@ mod tests {
         );
     }
 
-    /// The PHP instrumentation case: an empty string must be distinguishable
+    /// The PHP instrumentation case: `getHeaderLine('Content-Length')` sends ""
-    /// from any other unparsable value in the reported diagnostics.
+    /// when the header is absent. The span must be kept with a null attribute.
     #[test]
-    fn test_prepare_trace_column_rewrites_reports_empty_string_value() {
+    fn test_prepare_trace_column_rewrites_nulls_empty_string_value() {
-        let rows = vec![Row {
+        let mut rows = Rows {
-            values: vec![Value {
+            schema: vec![ColumnSchema {
-                value_data: Some(ValueData::StringValue(String::new())),
+                column_name: "span_attributes.http.response.body.size".to_string(),
+                datatype: ColumnDataType::String as i32,
+                ..Default::default()
             }],
-        }];
+            rows: vec![Row {
+                values: vec![Value {
+                    value_data: Some(ValueData::StringValue(String::new())),
+                }],
+            }],
+        };
         let pending_rewrites = vec![PendingTraceColumnRewrite {
             col_idx: 0,
             target_type: ColumnDataType::Int64,
             column_name: "span_attributes.http.response.body.size".to_string(),
         }];
 
-        let err = prepare_trace_column_rewrites(&rows, pending_rewrites, "opentelemetry_traces")
+        prepare_trace_column_rewrites(&rows.rows, pending_rewrites, "opentelemetry_traces")
-            .unwrap_err();
+            .unwrap()
-        assert!(
+            .apply(&mut rows);
-            err.error.to_string().contains(
+
-                "'span_attributes.http.response.body.size' in table 'opentelemetry_traces' \
+        assert_eq!(rows.schema[0].datatype, ColumnDataType::Int64 as i32);
-                 from String(\"\") to Int64"
+        assert_eq!(rows.rows[0].values[0].value_data, None);
-            ),
-            "unexpected error message: {}",
-            err.error
-        );
     }
 
     #[test]

src/servers/src/otlp/coerce.rs

@@ -30,6 +30,9 @@ pub enum TraceCoerceError {
 // - String to Int64
 // - String to Float64
 // - String to Boolean
+// The empty string and "-" coerce to null instead of failing: instrumentations
+// fill numeric attributes from absent headers this way (PSR-7 `getHeaderLine`
+// returns ""), and rejecting the value would reject the whole span.
 //
 // Lossless signed-to-unsigned integer casts. These let the built-in data
 // models move from unsigned to signed integers while existing unsigned tables
@@ -86,6 +89,13 @@ pub fn coerce_value_data(
         return Ok(None);
     };
 
+    if let ValueData::StringValue(s) = v
+        && matches!(s.as_str(), "" | "-")
+        && is_supported_trace_coercion(request_type, target)
+    {
+        return Ok(None);
+    }
+
     let Some(value) = coerce_non_null_value(target, request_type, v) else {
         return Err(TraceCoerceError::Unsupported);
     };
@@ -273,6 +283,24 @@ mod tests {
         assert_eq!(result, Err(TraceCoerceError::Unsupported));
     }
 
+    #[test]
+    fn test_coerce_absent_value_strings_to_null() {
+        for s in ["", "-"] {
+            for target in [
+                ColumnDataType::Int64,
+                ColumnDataType::Float64,
+                ColumnDataType::Boolean,
+            ] {
+                let result = coerce_value_data(
+                    &Some(ValueData::StringValue(s.to_string())),
+                    target,
+                    ColumnDataType::String,
+                );
+                assert_eq!(result, Ok(None), "{s:?} to {target:?}");
+            }
+        }
+    }
+
     #[test]
     fn test_coerce_float64_to_int64_not_supported() {
         let result = coerce_value_data(