Warnings and errors¶

All live in tabsight.warnings.

Warnings¶

tabsight warns rather than stops (rule 3). Every warning is a subclass of TabsightWarning, so the standard library can filter them:

Warning	When
`TabsightWarning`	Base class
`DataWarning`	Something in the data may mislead (mixed currencies, summing IDs, % of total of an average), or a dashboard was shrunk to fit the size limit (see `dash.size_steps`)
`ProfileWarning`	A guess was made (assumed date order, guessed file encoding)
`CleanWarning`	Something about cleaning (a saved step for a missing column)
`FormatWarning`	A formatting fallback (unknown locale)

import warnings

import tabsight as ts

warnings.simplefilter("ignore", ts.warnings.ProfileWarning)  # hide one kind
warnings.simplefilter("default", ts.warnings.TabsightWarning)  # back to normal

To turn every tabsight warning into an error, use strict mode: ts.set_defaults(strict=True) or with ts.config(strict=True):.

Errors¶

Error	When
`TabsightError`	Base class (and strict-mode warnings)
`InvalidSpecError`	An argument, setting, spec, schema or saved decision failed validation (also a `ValueError`)
`ChartLintError`	`fig.lint(strict=True)` found a chart design problem (see lint rules)
`UnsafeInputError`	Input refused for safety: a URL or network-share path, an unsupported file type, Excel without defusedxml, an Excel decompression bomb, JSON nested too deeply, old pyarrow (also a `ValueError`)
`CalcError`	A formula cannot be parsed or evaluated (in `tabsight.formulas`; a subclass of `InvalidSpecError`)
`BadRequest`	A server-mode request from the page is malformed (in `tabsight.server`; a subclass of `InvalidSpecError`). The web helpers answer it with HTTP 400

Messages show untrusted values shortened and quoted, so data cannot flood or fake error messages.