Map scraped UB-TUM branch hours onto NavigaTUM entry ids via `ub_tum_libraries.csv`.

Pick the German variant from a TranslatableStr-shaped dict; pass-through otherwise.

Extract the building prefix from the building_ids

Extract the name and the possible short_name

Reconstruct a full areatree line `<building_ids>:<name>:<id>[,<visible_id>]`.

Build an Iris roster frame; gebaeude_code is the `@`-suffix, as Iris reports it.

Split out so tests can pin `last_update` instead of reading the wall clock.

Freeze the visible count of a higher-priority section the first time a lower-priority hit lands. No-op if the section is empty (it stays at 0) or alre

Attempts to load a file from the local filesystem first, falling back to downloading it via HTTP. This allows users to provide static files in the Do

Map scraped canteen hours onto NavigaTUM entry ids as `OpeningHoursSchema` records. Canteens are matched to entries via the hand-authored `m

Return the current+next-ISO-week menu for `canteen`, fetching from eat-api on a cache miss. `today` is injected rather than read from the clock so th

Parse the filename of an image to get the id and index. Only the trailing `_<index>` is split off, so ids may themselves contain undersc

Set up the loglevels with colors, with correct terminal colors

Reconstruct the nested API dict from flat DataFrame columns.

Run ``add_aliases`` on a single synthetic row and return the resulting alias list.

Left-join patch_df onto df and coalesce so patch values win over originals. New rows from patch_df (ids not in df) are also added.

Extract the lines from the areatree file via a generator pattern Ignores: - Empty lines, - comment lines and - comments in lines

Build a two-canteen feed frame matching `StudierendenwerkSchema`.

Build a one-row frame from the valid baseline, overriding named columns.

Whether `schedule` carries any `lecture:`/`break:` macro.

Whether `schedule` carries any `PH` (public-holiday) selector.

Derive the NavigaTUM building ids that have at least one Iris-matched learning room. Each Iris room's `raum_nr_architekt` (the raw `<arch_na

The full document shape the data pipeline exports for one `events.csv` row.

Insert a new node under `parent_id`, sorted alphabetically among existing siblings to keep diffs against hand-edited entries minimal.

Zero-padded `YYYY-MM-DD` and a real calendar date (rejects e.g. `2026-13-40`).

Load the TUMonline organisation catalogue for `lang`. Dtypes enforced by `OrgsSchema`.

Narrow a `SemesterSchema` cell to `date` for the type checker (it always is).

Collapse the lookup frame into a plain dict for ergonomic assertions.

One node as `generate_buildings_overview` expects it in the global lookup.

Build the (id, type, short_name, parents) frame the lookup consumes.

Read a CSV against a dataframely schema, with whitespace stripping. Narrowed dtypes (`pl.Enum`, `pl.Categorical`) are read as `pl.String` fi

Build a one-row frame from the valid baseline, overriding named columns.

Split a line from the areatree file into the three parts (building-ids,name,internal-id) The syntax is building-id(s):name(s):internal-id[,v

For each entry, the code-like short_name that prefixes its rooms' arch names. Walks ``[self, immediate_parent, ..., root]`` nearest-first an

Add ``col_name`` as a typed null column if absent.

Add any of ``columns`` that are absent as typed null columns.

Extract the item that will be finally exported to the api