Feat/next dose

[TiaTuinstra]

Add next dose number to overdue list and validity markers to immunization history chart (Optional flags)

Summary

• Next dose number: The overdue/due vaccines list now displays the upcoming dose number for each immunization (e.g. "Varicella (2nd dose)" instead of "Varicella - 2"), controlled by the new preprocess.include_dose config flag (default: false).
• Validity markers: The immunization history chart now optionally renders distinct symbols for valid (⬤) and invalid (○) rather than a single filled dot for all received immunizations, controlled by preprocess.show_validity_markers (default: false).
  • Validity suffixes (- Valid / - Invalid) on received doses are parsed and surfaced in the notice; raises ValueError on mixed datasets (some doses have markers, others don't) to prevent misleading output

Changes

pipeline/preprocess.py

• format_vaccine_due_list() — formats due-vaccine strings from "Name - N" → "Name (Nth dose)" with correct ordinal suffixes
• normalize_validity_status() / collapse_validity_statuses() — strict normalization of validity tokens with precedence rules (unknown > mixed > valid > invalid)
• classify_dataset_validity() — dataset-level pre-scan returning all_present / all_absent / mixed before per-client processing begins
• build_received_rows() — row construction with validity tracking, deduplication, and row-splitting for dates where the same vaccine appears as both valid and invalid
• build_preprocess_result() — dataset-level validity classification, conditional error/warning branching, and dose formatting wired in

templates/conf.typ

• Replaces single vax symbol with vax_valid/vax_invalid
• New validity-dot() function dispatches the correct glyph based on validity status
• immunization-table() now accepts and threads show_validity_markers through cell rendering

templates/en_template.py / fr_template.py

• Pass SHOW_VALIDITY_MARKERS placeholder into Typst context; resolved at render time from the pipeline config

pipeline/generate_notices.py

• Reads preprocess.show_validity_markers from config and passes it to build_template_context

config/parameters.yaml

• Added preprocess.include_dose: false and preprocess.show_validity_markers: false under new preprocess key

Validity Parsing Logic

Received dose data arrives in IMMS GIVEN as semicolon-delimited segments, each optionally suffixed with - Valid or - Invalid (e.g. May 1, 2020 - DTaP - Valid; Jun 15, 2021 - MMR). The new parsing logic processes this in four stages:

1. Dataset-level pre-scan (classify_dataset_validity)
Before any per-client work, the full IMMS GIVEN column is scanned to classify whether validity suffixes are all_present, all_absent, or mixed across the dataset. This drives the error/warning decision upfront rather than accumulating ambiguous per-record unknowns.

2. Segment parsing (parse_dose_segments)
Each segment is matched against a regex that captures date, vaccine name, and an optional Valid/Invalid suffix. Unrecognised suffixes, typos, or missing suffixes all normalize to "unknown" via a strict two-value allowlist — only the exact casings "valid", "Valid", "invalid", "Invalid" are accepted.

3. Same-vaccine deduplication (_deduplicate_vaccines_for_date)
If the same vaccine appears multiple times on one date (e.g. from duplicate records), their validity statuses are collapsed: unknown dominates (data quality signal), then valid, then invalid.

4. Row construction and splitting (build_received_rows)
Doses are grouped by date. If a single date has a mix of valid and invalid vaccines across different diseases, that date is split into two display rows — one for valid entries, one for invalid — so no single table cell carries a misleading combined status. Rowspan values are computed accordingly for template rendering.

[jangevaare]

Looks good to me. AI had some small suggestions that I added to the branch with associated tests. The main one in my mind is the localization support for the dose labels.

restored the encryption default, propagated custom configuration correctly, localized dose labels, supported lowercase validity suffixes, documented intended validity grouping, and handled empty dose suffixes safely

[jangevaare]

There was a little bit more that came out in e2e tests to fix.

default french template spans 3 pages (unrelated), but warned and some reason accepted in our tests.

Also sorting out logic for combinations of dose label desired (config - yes/no), versus dose number bearing overdue disease list schema used (yes/no), and behaviour when this schema is used but dose number is blank.

[jangevaare]

Closes #146