fix(data-format): preserve values when expanding scientific notation

backend/common/utils/data_format.py

@@ -55,7 +55,7 @@ class DataFormat:
             """格式化浮点数，避免科学记数法"""
             if value == 0:
                 return "0"
-            formatted = str(Decimal(str(value)))
+            formatted = format(Decimal(str(value)), 'f')
             if '.' in formatted:
                 formatted = formatted.rstrip('0').rstrip('.')
             return formatted

backend/tests/test_data_format.py

@@ -0,0 +1,94 @@
+"""Regression tests for numeric values shared by query results and Excel exports."""
+
+import ast
+from decimal import Decimal
+from pathlib import Path
+
+import pytest
+
+
+@pytest.fixture(scope="module")
+def convert_numbers():
+    # Follow the existing isolated tests without importing the chat/LLM stack.
+    source = Path(__file__).resolve().parents[1] / "common/utils/data_format.py"
+    tree = ast.parse(source.read_text(encoding="utf-8"))
+    node = next(
+        node for node in tree.body if getattr(node, "name", None) == "DataFormat"
+    )
+    namespace = {"Decimal": Decimal}
+    exec(
+        compile(ast.Module(body=[node], type_ignores=[]), str(source), "exec"),
+        namespace,
+    )
+    return namespace["DataFormat"].convert_large_numbers_in_object_array
+
+
+@pytest.mark.parametrize(
+    ("value", "expected"),
+    [
+        (1.23e20, "123000000000000000000"),
+        (-1.23e20, "-123000000000000000000"),
+        (1.23e-10, "0.000000000123"),
+        (-1.23e-10, "-0.000000000123"),
+        (1e20, "100000000000000000000"),
+        (1e-7, "0.0000001"),
+        (1e10, "10000000000"),
+        (12345678901.25, "12345678901.25"),
+        (0.0, "0"),
+        (-0.0, "0"),
+    ],
+)
+def test_float_conversion_preserves_value_without_scientific_notation(
+    convert_numbers, value, expected
+):
+    assert convert_numbers([{"value": value}]) == [{"value": expected}]
+
+
+def test_values_outside_conversion_thresholds_keep_their_types(convert_numbers):
+    row = {
+        "regular_float": 123.4,
+        "small_float_boundary": 1e-6,
+        "below_float_threshold": 9999999999.5,
+        "below_int_threshold": 999999999999999,
+        "large_integer": 1000000000000000,
+        "negative_integer": -1000000000000000,
+        "zero": 0,
+        "flag": True,
+        "text": "1.23e20",
+        "missing": None,
+    }
+
+    converted = convert_numbers([row])[0]
+
+    assert converted == {
+        **row,
+        "large_integer": "1000000000000000",
+        "negative_integer": "-1000000000000000",
+    }
+    for key in ("regular_float", "small_float_boundary", "below_float_threshold"):
+        assert isinstance(converted[key], float)
+
+
+def test_scientific_notation_is_converted_in_nested_rows(convert_numbers):
+    rows = [{"metrics": {"amount": 1.23e20}, "series": [{"value": 1.23e-10}]}]
+
+    assert convert_numbers(rows) == [
+        {
+            "metrics": {"amount": "123000000000000000000"},
+            "series": [{"value": "0.000000000123"}],
+        }
+    ]
+    assert rows[0]["metrics"]["amount"] == 1.23e20
+
+
+def test_excel_integer_threshold_remains_supported(convert_numbers):
+    assert convert_numbers(
+        [{"below": 99999999999, "boundary": 100000000000, "amount": 1.23e20}],
+        int_threshold=1e11,
+    ) == [
+        {
+            "below": 99999999999,
+            "boundary": "100000000000",
+            "amount": "123000000000000000000",
+        }
+    ]