Bridge USD and EUR across all vintages at a chosen exchange-rate period

[Wegatriespython]

This makes a configured currency bridge usable across all deflator vintages rather than only the one it was configured at, and adds the deflator data needed to reach recent years. Follow-up to #61; continues #25. The bridge-year interpretation follows @khaeru's proposal discussed in iiasa/sparccle-workflow#19.

configure_currency(method, period) previously redefined only the {currency}_{period} unit in terms of USD. Because each vintage is defined relative to its own anchor (EUR_2005, USD_2005), redefining a single non-anchor vintage detached it from the rest of the chain: configure_currency("EXC", 2010) could convert USD_2010 → EUR_2010 but raised DimensionalityError for EUR_2024. Only period=2005 worked, because there the redefined unit is the anchor.

This PR reinterprets period as the exchange-rate bridge year only:

USD_y --(USD deflator)--> USD_period --(exchange rate @ period)--> EUR_period --(EUR deflator)--> EUR_z

• configure_currency() anchors the bridge at the target currency's chain anchor ({currency}_2005), scaled by the deflator factor between period and 2005, so the whole vintage chain stays connected. period=2005 results are unchanged.
• A registry holds one bridge per target currency: re-running with the same (method, period) is a no-op; changing either raises ValueError rather than silently re-anchoring earlier conversions.
• Adds USD_2023, USD_2024 and EUR_2015 to definitions.txt.

To write that bridge, the EUR deflator factors are now named in definitions.txt : _EUR_deflator_2010 = 0.901141, with EUR_2010 = _EUR_deflator_2010 * EUR_2005 as before, so configure_currency() can define the bridge as a plain pint expression, EUR_2005 = USD_2010 / 0.754309 / _EUR_deflator_2010, and let pint do the arithmetic. The factors follow the underscore prefix as _gwp, since they are not units. DEVELOPING.rst records that a new period or currency now needs a matching factor and chain anchor added by hand.

Data sourcing: the USD factors are each year's World Bank US GDP deflator (NY.GDP.DEFL.ZS, series last updated 2026-07-13) divided by the 2015 reference value and rounded to 4 decimals, following the existing 2000–2022 entries; EUR_2015 uses the euro-area series (NY.GDP.DEFL.KD.ZG, aggregate XC), taken from the 2026-04-08 vintage.

Adds test_currency_bridge covering every supported period and every defined EUR vintage.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.72%. Comparing base (3b78494) to head (cf9cb50).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #64      +/-   ##
==========================================
+ Coverage   91.48%   91.72%   +0.23%     
==========================================
  Files           6        6              
  Lines         282      290       +8     
==========================================
+ Hits          258      266       +8     
  Misses         24       24              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[khaeru]

Hi, thanks for a PR. Can you clarify whether you used LLMs in preparing this, and how?

[Wegatriespython]

Hi, yes an LLM—Claude Code, was used to prepare this PR, upon request from the SPARCCLE workflow team to support alternate exchange rate periods (2024). When trying to use non-2005 periods, DimensonalityError came up.

I reviewed the code, however since I am not an expert on this repo, the review was high level.

[Wegatriespython]

So I have force-pushed an update to slim down this PR, changes to updater have been cut out and deferred to a future PR. This PR is now only fixing the bridge bug plus adds data for USD_2023,USD_2024, EUR_2015) needed to use recent periods.

Hopefully this is in a more reviewable state now.

[byersiiasa]

Hi @khaeru - can we please get this merged or is there something blocking it?
We need it on SPARCCLE scenario explorer workflow so we can actually do analysis. Thanks

[Wegatriespython]

Will rebase and reduce some tech debt with the cache workaround in a bit. WB also put out new values for US deflators, will rope those in as well.

[Wegatriespython]

Pushed the rebase on main with the following changes :

1. The earlier cache workaround via registry._build_cache() is gone, it was there earlier because configure_currency() read the deflator factor by converting EUR_period to EUR_2005, which cached those units' pre-bridge dimensionality immediately before define() made it stale. With the factors now moved to definitions.txt, the bridge is a pint expression that pint resolves.
2. USD_2023 and USD_2024 are refreshed to the current World Bank vintage, 1.2577 and 1.2889, replacing the 1.2431 and 1.2739 first pushed here.

[byersiiasa]

Hi - can we please merge it now or please have some quick feedback?

This is getting embarrassing for the project that it takes us 3 months to update currency conversion.