sadpandajoe commented on code in PR #43770:
URL: https://github.com/apache/superset/pull/43770#discussion_r4234908254
##########
superset/mcp_service/chart/schemas.py:
##########
@@ -3939,6 +4058,477 @@ def validate_unique_column_labels(self) ->
"XYChartConfig":
return self
+class BulletChartConfig(BaseChartConfig):
+ """Config for bullet charts (viz_type ``bullet``)."""
+
+ # Semantic field names are exposed to MCP clients; validation aliases and
the
+ # native adapter accept saved Explore ``form_data`` without weakening the
+ # unknown-field checks that catch misspelled controls.
+ model_config = ConfigDict(extra="ignore", populate_by_name=True)
+
+ # Saved query roles are internal sort context, never typed authoring input.
+ # Keeping them out of dimensions preserves omission and native SQL shapes.
+ _inherited_groupby: list[str | dict[str, Any]] | None =
PrivateAttr(default=None)
+
+ @property
+ def order_dimensions(self) -> Sequence[ColumnRef | str | dict[str, Any]]:
+ """Return authored dimensions or the inherited native sort
hierarchy."""
+ if self.dimensions is None and self._inherited_groupby is not None:
+ return self._inherited_groupby
+ return self.dimensions or []
+
+ chart_type: Literal["bullet"] = "bullet"
+ metric: ColumnRef = Field(
+ ...,
+ description="Numeric measure shown by each bullet bar",
+ )
+ dimensions: List[ColumnRef] | None = Field(
+ None,
+ validation_alias=AliasChoices("dimensions", "groupby"),
+ description=(
+ "Category hierarchy; one bullet row per unique combination. Omit
to "
+ "keep the saved hierarchy on update; [] clears it."
+ ),
+ max_length=20,
+ )
+ filters: List[FilterConfig] | None = Field(None, max_length=100)
+ time_range: str | None = Field(
+ None,
+ min_length=1,
+ max_length=1000,
+ description=(
+ "Superset time range, e.g. 'Last 30 days' or '2025-01-01 :
2025-12-31'"
+ ),
+ )
+ row_limit: int = Field(
+ 10000,
+ ge=1,
+ le=50000,
+ description="Maximum bullet rows",
+ )
+ order_by: List[SortByConfig] = Field(
+ default_factory=list,
+ validation_alias=AliasChoices("order_by", "orderby", "order_by_cols"),
Review Comment:
`order_by` is declared with `AliasChoices("order_by", "orderby",
"order_by_cols")`, and Pydantic keeps only the first alias that is present.
Native Explore form data (and `_normalize_bullet_query_aliases`) concatenates
`orderby` and `order_by_cols`, so a native config such as `metric: "sales"`,
`groupby: ["region"]`, `orderby: []`, `order_by_cols: ['["sales", false]']`,
`row_limit: 1` loses the descending-sales sort here and may return a
lower-sales region. Should `adapt_native_form_data` merge these aliases into
one list before field selection, the way the saved-form-data path does?
##########
superset/mcp_service/chart/schemas.py:
##########
@@ -3939,6 +4058,477 @@ def validate_unique_column_labels(self) ->
"XYChartConfig":
return self
+class BulletChartConfig(BaseChartConfig):
+ """Config for bullet charts (viz_type ``bullet``)."""
+
+ # Semantic field names are exposed to MCP clients; validation aliases and
the
+ # native adapter accept saved Explore ``form_data`` without weakening the
+ # unknown-field checks that catch misspelled controls.
+ model_config = ConfigDict(extra="ignore", populate_by_name=True)
+
+ # Saved query roles are internal sort context, never typed authoring input.
+ # Keeping them out of dimensions preserves omission and native SQL shapes.
+ _inherited_groupby: list[str | dict[str, Any]] | None =
PrivateAttr(default=None)
+
+ @property
+ def order_dimensions(self) -> Sequence[ColumnRef | str | dict[str, Any]]:
+ """Return authored dimensions or the inherited native sort
hierarchy."""
+ if self.dimensions is None and self._inherited_groupby is not None:
+ return self._inherited_groupby
+ return self.dimensions or []
+
+ chart_type: Literal["bullet"] = "bullet"
+ metric: ColumnRef = Field(
+ ...,
+ description="Numeric measure shown by each bullet bar",
+ )
+ dimensions: List[ColumnRef] | None = Field(
+ None,
+ validation_alias=AliasChoices("dimensions", "groupby"),
+ description=(
+ "Category hierarchy; one bullet row per unique combination. Omit
to "
+ "keep the saved hierarchy on update; [] clears it."
+ ),
+ max_length=20,
+ )
+ filters: List[FilterConfig] | None = Field(None, max_length=100)
+ time_range: str | None = Field(
+ None,
+ min_length=1,
+ max_length=1000,
+ description=(
+ "Superset time range, e.g. 'Last 30 days' or '2025-01-01 :
2025-12-31'"
+ ),
+ )
+ row_limit: int = Field(
+ 10000,
+ ge=1,
+ le=50000,
+ description="Maximum bullet rows",
+ )
+ order_by: List[SortByConfig] = Field(
+ default_factory=list,
+ validation_alias=AliasChoices("order_by", "orderby", "order_by_cols"),
+ max_length=20,
+ description="Row order by a dimension name or the metric output
label/name",
+ )
+
+ # Presentation fields map one-for-one onto Bullet/transformProps.ts
controls.
+ ranges: List[float] = Field(
+ default_factory=list,
+ max_length=100,
+ description="Qualitative range thresholds shaded behind the measure",
+ )
+ range_labels: List[str] = Field(
+ default_factory=list,
+ validation_alias=AliasChoices("range_labels", "rangeLabels"),
+ max_length=100,
+ )
+ markers: List[float] = Field(
+ default_factory=list,
+ max_length=100,
+ description="Target values drawn as point markers",
+ )
+ marker_labels: List[str] = Field(
+ default_factory=list,
+ validation_alias=AliasChoices("marker_labels", "markerLabels"),
+ max_length=100,
+ )
+ marker_lines: List[float] = Field(
+ default_factory=list,
+ validation_alias=AliasChoices("marker_lines", "markerLines"),
+ max_length=100,
+ description="Reference values drawn as vertical lines",
+ )
+ marker_line_labels: List[str] = Field(
+ default_factory=list,
+ validation_alias=AliasChoices("marker_line_labels",
"markerLineLabels"),
+ max_length=100,
+ )
+ y_axis_format: str = Field(
+ "SMART_NUMBER",
+ validation_alias=AliasChoices("y_axis_format", "yAxisFormat"),
+ max_length=100,
+ )
+ show_labels: bool = Field(
+ False,
+ validation_alias=AliasChoices("show_labels", "showLabels"),
+ )
+ show_legend: bool = Field(
+ False,
+ validation_alias=AliasChoices("show_legend", "showLegend"),
+ )
+
+ @staticmethod
+ def _adapt_native_metric(value: Any) -> Any:
+ """Translate QueryFormMetric shapes into the shared ColumnRef
contract."""
+ if isinstance(value, str):
+ return {"name": value, "saved_metric": True}
+ if not isinstance(value, dict):
+ return value
+ if "expressionType" not in value:
+ # QueryObject's documented legacy saved-metric representation is a
+ # label-only object. Keep this adapter deliberately narrow: objects
+ # carrying ad-hoc fields must declare expressionType explicitly,
and
+ # semantic ColumnRef objects continue through normal validation.
+ if set(value) == {"label"}:
+ label = value["label"]
+ if not isinstance(label, str) or not label or len(label) > 255:
+ raise ValueError(
+ "legacy saved metric label must be a non-empty string
of "
+ "at most 255 characters"
+ )
+ return {"name": label, "saved_metric": True}
+ return value
+ expression_type = value.get("expressionType")
+ if expression_type == "SQL":
+ return {
+ "sql_expression": value.get("sqlExpression"),
+ "label": value.get("label") or value.get("sqlExpression"),
+ }
+ if expression_type != "SIMPLE":
+ raise ValueError("metric.expressionType must be 'SIMPLE' or 'SQL'")
+ column = value.get("column")
+ if isinstance(column, dict):
+ name = column.get("column_name")
+ else:
+ name = column
+ return {
+ "name": name,
+ "aggregate": value.get("aggregate"),
+ "label": value.get("label"),
+ }
+
+ @staticmethod
+ def _canonical_dimension_alias(value: Any, field_name: str) -> list[str]:
+ """Canonicalize semantic/native dimension aliases for conflict
checks."""
+ if not isinstance(value, list):
+ raise ValueError(f"{field_name} must be an array")
+ canonical: list[str] = []
+ for index, item in enumerate(value):
+ name: str | None
+ if isinstance(item, str):
+ name = item
+ elif isinstance(item, ColumnRef):
+ name = item.name
+ elif isinstance(item, dict):
+ name = next(
+ (
+ item[key]
+ for key in ("name", "column_name", "column")
+ if isinstance(item.get(key), str)
+ ),
+ None,
+ )
+ else:
+ name = None
+ if not name:
+ raise ValueError(
+ f"{field_name}[{index}] must identify a physical column"
+ )
+ canonical.append(name)
+ return canonical
+
+ @staticmethod
+ def _adapt_native_order_by(value: Any) -> Any: # noqa: C901
+ if value is None:
+ return []
+ if not isinstance(value, list):
+ raise ValueError("order_by must be an array")
+ result: list[Any] = []
+ for index, entry in enumerate(value):
+ if isinstance(entry, str):
+ if len(entry) > 2000:
+ raise ValueError(f"order_by[{index}] is too long")
+ try:
+ entry = json.loads(entry)
+ except json.JSONDecodeError:
+ # A bare output/column name is the ergonomic typed form.
+ result.append({"column": entry, "ascending": False})
+ continue
+ if isinstance(entry, dict):
+ result.append(entry)
+ continue
+ if not isinstance(entry, (list, tuple)) or len(entry) != 2:
+ raise ValueError(
+ f"order_by[{index}] must be [column, ascending_boolean]"
+ )
+ target, ascending = entry
+ if isinstance(target, dict):
+ target = target.get("label") or target.get("metric_name")
+ if not isinstance(target, str) or not target:
+ raise ValueError(f"order_by[{index}] needs a column or metric
label")
+ if not isinstance(ascending, bool):
+ raise ValueError(f"order_by[{index}] ascending value must be
boolean")
+ result.append({"column": target, "ascending": ascending})
+ return result
+
+ @staticmethod
+ def _adapt_native_filters(data: dict[str, Any]) -> None: # noqa: C901
+ if "adhoc_filters" not in data:
+ return
+ if "filters" in data:
+ raise ValueError("Use either filters or native adhoc_filters, not
both")
+ raw_filters = data.pop("adhoc_filters")
+ if not isinstance(raw_filters, list):
+ raise ValueError("adhoc_filters must be an array")
+ filters: list[dict[str, Any]] = []
+ temporal_pairs: list[tuple[str, str]] = []
+ inert_subjects: set[str] = set()
+ for index, raw_filter in enumerate(raw_filters):
+ if not isinstance(raw_filter, dict):
+ raise ValueError(f"adhoc_filters[{index}] must be an object")
+ if raw_filter.get("expressionType") != "SIMPLE":
+ raise ValueError(
+ f"adhoc_filters[{index}] must use expressionType='SIMPLE'"
+ )
+ if raw_filter.get("clause") not in (None, "WHERE"):
+ raise ValueError(f"adhoc_filters[{index}] must use
clause='WHERE'")
+ subject = raw_filter.get("subject")
+ operator = raw_filter.get("operator")
+ comparator = raw_filter.get("comparator")
+ if operator == "TEMPORAL_RANGE":
+ if not isinstance(subject, str) or not subject:
+ raise ValueError(
+ f"adhoc_filters[{index}] temporal filter needs subject"
+ )
+ if not isinstance(comparator, str) or not comparator:
+ raise ValueError(
+ f"adhoc_filters[{index}] temporal filter needs a range"
+ )
+ if comparator.casefold() == "no filter":
+ inert_subjects.add(subject)
+ else:
+ temporal_pairs.append((subject, comparator))
+ continue
+ if not isinstance(operator, str):
+ raise ValueError(f"adhoc_filters[{index}] needs an operator")
+ operator_map = {
+ "==": "=",
+ "EQUALS": "=",
+ "NOT_EQUALS": "!=",
+ "LESS_THAN": "<",
+ "LESS_THAN_OR_EQUAL": "<=",
+ "GREATER_THAN": ">",
+ "GREATER_THAN_OR_EQUAL": ">=",
+ "NOT_IN": "NOT IN",
+ "IS_NULL": "IS NULL",
+ "IS_NOT_NULL": "IS NOT NULL",
+ }
+ operator = operator_map.get(operator, operator)
+ filters.append({"column": subject, "op": operator, "value":
comparator})
+ if len(temporal_pairs) > 1:
+ raise ValueError(
+ "Multiple active native temporal filters cannot be represented
by "
+ "a single temporal_column/time_range pair"
+ )
+ if temporal_pairs:
+ subject, comparator = temporal_pairs[0]
+ if data.get("temporal_column") not in (None, subject) or data.get(
+ "time_range"
+ ) not in (None, "No filter", comparator):
+ raise ValueError(
+ "Native temporal filter conflicts with
temporal_column/time_range"
+ )
+ data["temporal_column"] = subject
+ data["time_range"] = comparator
+ elif len(inert_subjects) == 1:
+ data.setdefault("temporal_column", next(iter(inert_subjects)))
+ data["filters"] = filters
+
+ @model_validator(mode="before")
+ @classmethod
+ def adapt_native_form_data(cls, raw: Any) -> Any: # noqa: C901
+ """Accept recognized saved Bullet form_data and reject ambiguous
state."""
+ if not isinstance(raw, dict):
+ return raw
+ data = dict(raw)
+ # Null means no hierarchy was supplied, not a conflicting alias or a
+ # request to clear saved dimensions. Only an explicit [] clears them.
+ for key in ("dimensions", "groupby"):
+ if data.get(key) is None:
+ data.pop(key, None)
+ if "dimensions" in data and "groupby" in data:
+ dimensions = cls._canonical_dimension_alias(
+ data["dimensions"], "dimensions"
+ )
+ groupby = cls._canonical_dimension_alias(data["groupby"],
"groupby")
+ if dimensions != groupby:
+ raise ValueError(
+ "Conflicting Bullet dimension aliases: 'dimensions' and "
+ "native 'groupby' must identify the same physical columns
in "
+ "the same order; provide only one or make them equivalent"
+ )
+ # Avoid relying on AliasChoices precedence or JSON key order.
+ data.pop("groupby")
+ if data.get("viz_type") == "bullet":
Review Comment:
This `viz_type == "bullet"` branch never runs for
`generate_chart`/`update_chart` requests: `_normalize_chart_request_input` maps
`viz_type` to `chart_type` and pops `viz_type` before this validator sees the
config. So `{"config": {"viz_type": "bullet", "metric": "Revenue", "markers":
"0x64"}}` skips `_bullet_numeric_control_tokens` and fails strict float
validation, while the same dict validated directly against `BulletChartConfig`
is accepted with marker 100. Should the request normalizer carry native
provenance (as it does for `sunburst_v2`) so both entry points parse these
controls the same way?
##########
superset/mcp_service/chart/preview_utils.py:
##########
@@ -1809,7 +1815,1041 @@ def generate_xy_vega_lite_preview(
spec["transform"] = [
{"fold": series_columns, "as": [series_field, value_field]}
]
- encoding["y"] = {"field": value_field, "type": "quantitative"}
+ # Vega-Lite stacks bar and area marks colored by a nominal field
+ # by default; follow the chart's saved ``stack`` control instead.
+ encoding["y"] = {
+ "field": value_field,
+ "type": "quantitative",
+ "stack": _xy_vega_lite_stack(form_data.get("stack")),
+ }
encoding["color"] = {"field": series_field, "type": "nominal"}
return preview
+
+
+def _xy_vega_lite_stack(stack: Any) -> str | None:
+ """Map the ECharts timeseries ``stack`` control to a Vega-Lite stack."""
+ if not stack:
+ return None
+ return "normalize" if stack == "Expand" else "zero"
+
+
+def _truncate_utf8(value: str, max_bytes: int) -> str:
+ """Return bounded, replacement-decoded UTF-8 text.
+
+ Encoding even the non-truncated path is intentional: Python strings may
+ contain unpaired surrogates, while MCP/JSON responses must always be valid
+ UTF-8. Slicing by characters before encoding also prevents an
+ attacker-sized string from being encoded in full.
+ """
+ if max_bytes <= 0:
+ return ""
+ candidate = value[:max_bytes]
+ encoded = candidate.encode("utf-8", errors="replace")
+ if len(encoded) <= max_bytes and len(candidate) == len(value):
+ return encoded.decode("utf-8", errors="replace")
+ suffix = "... [truncated]"
+ suffix_bytes = suffix.encode()
+ if max_bytes <= len(suffix_bytes):
+ return suffix_bytes[:max_bytes].decode("ascii")
+ content_limit = max(0, max_bytes - len(suffix_bytes))
+ content = encoded[:content_limit].decode("utf-8", errors="ignore")
+ return content + suffix
+
+
+_MAX_BULLET_FIELDS = 256
+
+
+_MAX_BULLET_FIELD_BYTES = 1000
+
+
+_MAX_BULLET_TEXT_BYTES = 2000
+
+
+# ECMAScript WhiteSpace and LineTerminator characters used by trim/Number.
+_JAVASCRIPT_WHITESPACE = (
+ "\t\n\v\f\r \u00a0\u1680\u2000\u2001\u2002\u2003\u2004"
+ "\u2005\u2006\u2007\u2008\u2009\u200a\u2028\u2029\u202f\u205f\u3000\ufeff"
+)
+
+
+_MAX_BULLET_TOKENS = 256
+
+
+_ENUM_SCALAR_TYPES = (str, int, float, bool, Decimal)
+
+
+class BulletOutputError(ValueError):
+ """A Bullet query result cannot be rendered without guessing its roles."""
+
+ def __init__(self, message: str, error_type: str =
"MalformedBulletOutput") -> None:
+ super().__init__(message)
+ self.error_type = error_type
+
+
+@dataclass(frozen=True)
+class BulletRenderModel:
+ """Strict, frontend-aligned data and presentation roles for one preview."""
+
+ rows: list[dict[str, Any]]
+ metric_field: str
+ dimensions: list[str]
+ measures: list[float]
+ ranges: list[float]
+ range_labels: list[str]
+ markers: list[float]
+ marker_labels: list[str]
+ marker_lines: list[float]
+ marker_line_labels: list[str]
+ y_axis_format: str
+ show_labels: bool
+ show_legend: bool
+
+
+def _form_metric_label(metric: Any) -> str | None:
+ """Return the result-column label for a native QueryFormMetric."""
+ if type(metric) is str:
+ return metric
+ if type(metric) is not dict:
+ return None
+ if label := dict.get(metric, "label"):
+ return label if type(label) is str else None
+ if dict.get(metric, "expressionType") == "SQL":
+ expression = dict.get(metric, "sqlExpression")
+ return expression if type(expression) is str and expression else None
+ column = dict.get(metric, "column")
+ column_name = dict.get(column, "column_name") if type(column) is dict else
column
+ aggregate = dict.get(metric, "aggregate")
+ if type(column_name) is str and type(aggregate) is str:
+ return f"{aggregate}({column_name})"
+ return None
+
+
+def _form_column_label(column: Any) -> str | None:
+ """Return the result-column label for a native QueryFormColumn."""
+ if type(column) is str:
+ return column
+ if type(column) is not dict:
+ return None
+ # getColumnLabel: an adhoc column without a label is keyed by its SQL.
+ for key in ("label", "sqlExpression", "column_name"):
+ if type(value := dict.get(column, key)) is str and value:
+ return value
+ return None
+
+
+def _require_result_field(label: str | None, row: dict[str, Any], role: str)
-> str:
+ """Resolve a role without falling back to an unrelated result field."""
+ if not label:
+ raise BulletOutputError(f"Bullet {role} has no declared result alias")
+ if label in dict.keys(row):
+ return label
+ matches = sorted(
+ field
+ for field in dict.keys(row)
+ if type(field) is str and field.casefold() == label.casefold()
+ )
+ if len(matches) == 1:
+ return matches[0]
+ if matches:
+ raise BulletOutputError(
+ f"Bullet {role} alias {label!r} is ambiguous; candidates: "
+ f"{', '.join(matches)}"
+ )
+ raise BulletOutputError(
+ f"Bullet {role} alias {label!r} is missing from query output"
+ )
+
+
+def _safe_enum_backing(value: Any) -> Any:
+ """Extract Enum's stored value without public descriptors/conversions."""
+ value_type = type(value)
+ try:
+ mro = type.__getattribute__(value_type, "__mro__")
+ except (AttributeError, TypeError): # pragma: no cover - normal types
have MRO
+ return value
+ if type(mro) is not tuple or not any(base is Enum for base in mro):
+ return value
+ try:
+ backing = object.__getattribute__(value, "_value_")
+ except Exception as ex:
+ raise BulletOutputError("Bullet output contains an unreadable enum")
from ex
+ if not any(type(backing) is allowed for allowed in _ENUM_SCALAR_TYPES):
+ raise BulletOutputError("Bullet output contains an unsupported enum
value")
+ return backing
+
+
+def _decimal_javascript_string(value: Decimal) -> str:
+ """Render an exact binary64 spelling with JavaScript Number thresholds."""
+ sign, digits_tuple, exponent = Decimal.as_tuple(value)
+ if type(exponent) is not int: # finite Decimals always have an integer
exponent
+ raise BulletOutputError("Bullet dimension contains a non-finite
Decimal")
+ if not any(digits_tuple):
+ return "0"
+
+ digits = "".join(str(digit) for digit in digits_tuple)
+ adjusted = len(digits) + exponent - 1
+ prefix = "-" if sign else ""
+ if -6 <= adjusted < 21:
+ point = len(digits) + exponent
+ if point <= 0:
+ text = f"0.{('0' * -point)}{digits}"
+ elif point >= len(digits):
+ text = digits + ("0" * (point - len(digits)))
+ else:
+ text = f"{digits[:point]}.{digits[point:]}"
+ if "." in text:
+ text = text.rstrip("0").rstrip(".")
+ return prefix + text
+
+ fraction = digits[1:].rstrip("0")
+ coefficient = digits[0] + (f".{fraction}" if fraction else "")
+ exponent_text = f"+{adjusted}" if adjusted >= 0 else str(adjusted)
+ return f"{prefix}{coefficient}e{exponent_text}"
+
+
+def _javascript_number_string(value: int | float | Decimal) -> str:
+ """Apply JSON-number -> IEEE-754 Number -> JavaScript String semantics.
+
+ Exact result scalars can retain precision that the frontend cannot: JSON
+ parsing first rounds a numeric token to binary64, and ``String`` then emits
+ the shortest round-tripping decimal with fixed notation for exponents in
+ [-6, 20]. Converting exact builtin scalars to an exact builtin float keeps
+ the path hook-free. Python and JavaScript use the same shortest
+ round-tripping binary64 digits; ``_decimal_javascript_string`` only adjusts
+ the notation thresholds and exponent spelling.
+
+ A finite integer or Decimal outside binary64's range becomes an infinity
+ after JSON parsing, matching JavaScript. Non-finite source values are
+ rejected by the trusted scalar normalizer before this helper is called.
+ """
+ value_type = type(value)
+ if value_type not in {int, float, Decimal}:
+ raise BulletOutputError("Bullet dimension contains an unsupported
number")
+ if value_type is float and not math.isfinite(value):
+ raise BulletOutputError("Bullet dimension contains a non-finite
number")
+ if isinstance(value, Decimal) and not Decimal.is_finite(value):
+ raise BulletOutputError("Bullet dimension contains a non-finite
Decimal")
+ try:
+ number = float(value)
+ except OverflowError:
+ number = -math.inf if value < 0 else math.inf
+
+ if math.isinf(number):
+ return "-Infinity" if number < 0 else "Infinity"
+ if number == 0:
+ # String(-0) is "0" even though JSON.parse preserves negative zero.
+ return "0"
+ return _decimal_javascript_string(Decimal(float.__repr__(number)))
+
+
+def _bullet_category_value( # noqa: C901
+ value: Any, dimension: str, row_index: int
+) -> tuple[Any, str]:
+ """Return a JSON-safe value and bounded frontend ``String(value)`` text.
+
+ The trusted scalar normalizer is type-exact and does not dispatch through
+ application hooks. Vega data retains the normalized Chart Data wire value
+ (including epoch-ms temporal numbers); only the derived category key and
+ ASCII label use the JavaScript-compatible text.
+ """
+ from superset.mcp_service.chart.query_result import (
+ _bounded_utf8_length,
+ _chart_data_duration_text,
+ _chart_data_temporal_number,
+ _is_chart_data_duration_scalar,
+ _is_chart_data_temporal_scalar,
+ _normalize_scalar,
+ )
+
+ normalized: Any
+ reason: str | None
+ if type(value) in {list, tuple, dict}:
+ data, failure = first_query_data(
+ {"queries": [{"data": [{"value": value}]}]},
temporal_json_numbers=True
+ )
+ if failure is not None or data is None:
+ raise BulletOutputError(
+ f"Bullet dimension {dimension!r} row {row_index} "
+ "has an invalid or unbounded container value"
+ )
+ normalized = data[0]["value"]
+ reason = None
+ elif _is_chart_data_temporal_scalar(value := _safe_enum_backing(value)):
+ normalized, reason = _chart_data_temporal_number(value)
+ elif _is_chart_data_duration_scalar(value):
+ normalized, reason = _chart_data_duration_text(value)
+ else:
+ normalized, reason = _normalize_scalar(value)
+ if (
+ reason is None
+ and normalized is None
+ and value is not None
+ and type(value) in {float, np.float16, np.float32, np.float64}
+ and math.isinf(float(value))
+ ):
+ # NaN keeps the shared null normalization, but an infinite source
+ # value has no faithful category text and is rejected.
+ reason = "contains a non-finite number"
+ if (
+ reason is None
+ and type(normalized) is str
+ and _bounded_utf8_length(normalized, _MAX_BULLET_TEXT_BYTES) is
None
+ ):
+ reason = "exceeds the size limit"
+ if reason is not None:
+ if reason == "an unsupported or subclassed value":
+ reason = "has an unsupported value type"
+ elif not reason.startswith(("contains ", "exceeds ", "has ")):
+ reason = f"contains {reason}"
+ raise BulletOutputError(
+ f"Bullet dimension {dimension!r} row {row_index} {reason}"
+ )
+
+ value_type = type(normalized)
+ if value_type in {list, dict}:
+ text = _bullet_container_category_text(normalized, dimension,
row_index)
+ elif normalized is None:
+ text = "null"
+ elif value_type is str:
+ text = normalized
+ elif value_type is bool:
+ text = "true" if normalized else "false"
+ elif value_type is int or value_type is float or value_type is Decimal:
+ text = _javascript_number_string(normalized)
+ else:
+ raise BulletOutputError(
+ f"Bullet dimension {dimension!r} row {row_index} has an "
+ "unsupported value type"
+ )
+
+ if _bounded_utf8_length(text, _MAX_BULLET_TEXT_BYTES) is None:
+ raise BulletOutputError(
+ f"Bullet dimension {dimension!r} row {row_index} exceeds the size
limit"
+ )
+ return normalized, text
+
+
+def _bullet_container_category_text(value: Any, dimension: str, row_index:
int) -> str:
+ """Stringify validated containers like JavaScript with a bounded text
budget."""
+ from superset.mcp_service.chart.query_result import _bounded_utf8_length
+
+ if type(value) is dict:
+ return "[object Object]"
+ parts: list[str] = []
+ size = 0
+ for index, item in enumerate(value):
+ if type(item) in {list, dict}:
+ text = _bullet_container_category_text(item, dimension, row_index)
+ else:
+ text = (
+ ""
+ if item is None
+ else _bullet_category_value(item, dimension, row_index)[1]
+ )
+ text_size = _bounded_utf8_length(text, _MAX_BULLET_TEXT_BYTES)
+ size += (text_size if text_size is not None else
_MAX_BULLET_TEXT_BYTES + 1) + (
+ index > 0
+ )
+ if size > _MAX_BULLET_TEXT_BYTES:
+ raise BulletOutputError(
+ f"Bullet dimension {dimension!r} row {row_index} exceeds the
size limit"
+ )
+ parts.append(text)
+ return ",".join(parts)
+
+
+def _javascript_numeric_string(value: str) -> float:
+ """Parse a nonempty trimmed string using JavaScript Number's grammar."""
+ if re.fullmatch(r"0[xX][0-9a-fA-F]+|0[bB][01]+|0[oO][0-7]+", value):
+ return float(int(value, 0))
+ if re.fullmatch(
+
r"[+-]?(?:Infinity|(?:[0-9]+(?:\.[0-9]*)?|\.[0-9]+)(?:[eE][+-]?[0-9]+)?)",
+ value,
+ ):
+ return float(value)
+ raise ValueError("Invalid JavaScript number spelling")
+
+
+def _bullet_number(value: Any, row_index: int, metric_field: str) -> float:
+ """Apply the frontend's useful ``Number(value ?? 0)`` numeric subset."""
+ value = _safe_enum_backing(value)
+ if value is None:
+ number = 0.0
+ elif type(value) is bool:
+ raise BulletOutputError(
+ f"Bullet metric {metric_field!r} row {row_index} returned a
boolean"
+ )
+ elif type(value) is int or type(value) is float or type(value) is Decimal:
+ try:
+ number = float(value)
+ except (TypeError, ValueError, OverflowError) as ex:
+ raise BulletOutputError(
+ f"Bullet metric {metric_field!r} row {row_index} is not
numeric"
+ ) from ex
+ elif type(value) is str:
+ if len(value) > _MAX_BULLET_TEXT_BYTES:
+ raise BulletOutputError(
+ f"Bullet metric {metric_field!r} row {row_index} is not
numeric"
+ )
+ stripped = value.strip(_JAVASCRIPT_WHITESPACE)
+ if not stripped:
+ raise BulletOutputError(
+ f"Bullet metric {metric_field!r} row {row_index} is not
numeric"
+ )
+ try:
+ number = _javascript_numeric_string(stripped)
+ except (ValueError, OverflowError) as ex:
+ raise BulletOutputError(
+ f"Bullet metric {metric_field!r} row {row_index} returned "
+ f"non-numeric text"
+ ) from ex
+ else:
+ raise BulletOutputError(
+ f"Bullet metric {metric_field!r} row {row_index} is not numeric"
+ )
+ if not math.isfinite(number):
+ raise BulletOutputError(
+ f"Bullet metric {metric_field!r} row {row_index} is NaN or
infinite"
+ )
+ return number
+
+
+def _bullet_string_tokens(value: Any) -> list[str]:
+ """Parse labels exactly like the frontend's comma tokenizer."""
+ value = _safe_enum_backing(value)
+ if value is None:
+ return []
+ if type(value) is not str or len(value) > _MAX_BULLET_TEXT_BYTES:
Review Comment:
The schema lets `range_labels`/`marker_labels`/`marker_line_labels` carry up
to 100 labels of 200 characters each, but this check caps the whole
comma-joined string at `_MAX_BULLET_TEXT_BYTES` (2000). A config that passes
validation, e.g. 21 range labels of 100 characters, gets joined into one
control string of about 2,100 characters, so the render model raises
`BulletOutputError` and `generate_chart`, saved-chart reads and exports fail
with a malformed-output error even though the query returned valid numbers.
Should the schema bound the total joined length, or this limit be raised to
match the schema's per-label/count limits?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
