Validators overview

Validators check a DataFrame without changing it. validate(df) returns a ValidationResult: errors for data that can’t be used, warnings for data that is suspicious. Invalid settings raise ValueError when the validator is created.

Which Validator Catches What

Problem	Validator
A local day with the wrong number of rows (DST days, naive wall-clock index)	DSTStructureValidator
Duplicate or unsorted timestamps, missing timestamps, mixed resolutions, one column missing for a stretch on a complete index, a late start or early end, too many nulls in the columns the model uses	CoverageValidator
Prices outside the SDAC limits, zero load, frozen values	PlausibilityValidator
Summary statistics for a first look	EdaValidator

Data gaps are easiest to see before transformers run: resampling with interpolation or forward fill hides them. Validators that work on raw source data (UTC index) take a tz parameter where local days matter.

ValidationResult

from epftoolbox2.data.validators import ValidationResult

@dataclass
class ValidationResult:
    is_valid: bool = True
    errors: List[str] = []
    warnings: List[str] = []
    info: Dict[str, Any] = {}          # counts, timestamps and details per check
    stats: pd.DataFrame = pd.DataFrame()  # a table per column or per day, depending on the validator
    validator: str = ""                 # class name of the validator that produced it

• bool(result) is result.is_valid.
• print(result) shows the status, errors, warnings, info (long lists shortened) and the size of stats.
• result.add_error(message) appends an error and marks the result invalid (useful in custom validators).
• result.to_frame() returns the messages as a table.

Printing Several Results as a Table

from epftoolbox2.data.validators import (
    CoverageValidator,
    DSTStructureValidator,
    results_to_frame,
)

validators = [DSTStructureValidator(), CoverageValidator(columns=["price"], min_coverage=1.0)]
results = [v.validate(df) for v in validators]
print(results_to_frame(results).to_string(index=False))

            validator  valid   level  message
DSTStructureValidator   True      ok
    CoverageValidator  False   error  Column 'price' has values at 168 of 240 expected 1h timestamps ...

results_to_frame takes a list of results, a list of (name, result) pairs or a dict {name: result}, and returns one row per error and warning (columns validator, valid, level, message). A result without messages gets one ok row.