Settings¶

Most settings are things you'd otherwise repeat on every call: your locale, your currency, whether your dates are day-first. There are two ways to set them.

For the whole program¶

Put this near the top of a script or notebook:

import tabsight as ts

ts.set_defaults(locale="en_IE", currency="EUR")
print(ts.format_number(1234.5, "currency"))

Output

€1,235

For one block of code¶

Inside a web server, two requests can run at the same time for two different users. Changing global defaults there would let one user's settings leak into the other's report. Use a with block instead; it only affects the code inside it, in that thread or async task:

with ts.config(locale="de_DE"):
    print(ts.format_number(1234.5, "currency", decimals=2))
print(ts.format_number(1234.5, "currency", decimals=2))

Output

1.234,50 €
€1,234.50

Calling ts.config(...) without with does nothing, so tabsight warns you if you try. That's an easy mistake to make, and a silent no-op would be worse.

When the same setting is given in several places, the most specific wins: an argument to the call, then a with ts.config block, then ts.set_defaults, then the built-in default.

Every setting¶

Setting	Default	What it does
`locale`	`"en_US"`	Number and date formats (list)
`currency`	`"USD"`	3-letter currency code
`dayfirst`	`None`	Day/month order for ambiguous dates; `None` means infer, then month-first
`fiscal_start`	`1`	Month the fiscal year starts, e.g. `4` for April
`week_start`	`"monday"`	`"monday"`, `"sunday"` or `"saturday"`
`missing`	`"gap"`	Empty periods: `"gap"`, `"zero"` or `"connect"`
`tz`	`None`	Time zone that time-zone-aware dates are converted to before grouping, like `"Europe/Dublin"`. By default they're grouped in their own zone
`theme`	`"light"`	`"light"`, `"dark"` or `"print"`
`palette`	`"classic10"`	Colour palette
`top`	`10`	Default top N for charts
`strict`	`False`	Turn tabsight warnings into errors
`quiet`	`False`	Hide data warnings
`max_size_mb`	`5.0`	Largest data table a dashboard file may carry; bigger ones are shrunk in steps (see dashboards)
`min_group_size`	`None`	Hide groups with fewer rows than this
`csv_safe`	`True`	Protect CSV downloads from spreadsheet formula injection
`picture`	`True`	In notebooks, add a still picture under each live chart, so a saved notebook still shows the chart when it is reopened untrusted or viewed on GitHub. Needs the `static` extra (dukpy); without it nothing changes
`cdn`	`False`	In notebooks, load ECharts from jsDelivr (pinned by a hash) instead of putting 1 MB into every chart. Charts then need the internet to draw

Every value is checked when you set it. A typo gets a suggestion rather than being ignored:

from tabsight.warnings import InvalidSpecError

try:
    ts.set_defaults(curency="EUR")
except InvalidSpecError as err:
    print(err)

Output

unknown setting(s): 'curency' (did you mean 'currency'?)

Strict mode¶

In a scheduled job you may want a warning to stop the run rather than scroll past in a log. strict=True turns every tabsight warning into a TabsightError:

import pandas as pd

from tabsight.warnings import TabsightError

ambiguous = pd.DataFrame({"Date": ["01/02/2017", "03/04/2017"], "Sales": [1.0, 2.0]})
with ts.config(strict=True):
    try:
        ts.profile(ambiguous)
    except TabsightError as err:
        print(err)

Output

ProfileWarning: 'Date': day/month order assumed month-first (no value above 12 settles it; pass dayfirst=)