Crystal-Builder test-data inventory
Task 0.9.3 maps the data shipped in Crystal-Builder's tests/data/ to the xtalforge subtask that uses it. Only the items that a phase actually needs are copied, with provenance in tests/corpus/PROVENANCE.md or THIRD_PARTY.md. Items marked "not copied" below are represented by hand-written fixtures or by the documented subset instead.
| Item | Contents | Used by |
|---|---|---|
blocks/*.xyz | MOF connection-point fragments (U4Ph, UOMe, UPh) | MOF builder, Task 6.12 |
dftb/* | DFTB+ outputs: Hessian, modes, DOS, band lines, MD, relaxation | External engine parsers, Task 4.9; the Si SCC output is also in crates/xtalforge-ff/tests/fixtures/dftb/ |
eqeq/mof5_cod_authors_program.json | reference EQeq charges for MOF-5 | Charge methods, Task 4.4; copied to crates/xtalforge-ff/tests/fixtures/eqeq/ |
eqeq/nist_asd_ionization.csv.gz, pubchem_periodic_table.csv | ionization data and periodic-table parameters | QEq/EQeq parameters, Task 4.4 |
hkust1_wide_nodes.xyz | wide-node MOF test fragment | MOF builder, Task 6.12 |
pormake_upstream.json | recorded PORMAKE output | MOF builder golden data, Task 6.12 |
rcsr/2dall.txt.gz, rcsr/3dall.txt.gz | RCSR net snapshots | Topology and RCSR search, Tasks 5.6 and 5.7 |
rod_*.xyz | rod SBU fragments for polymer/MOF assembly | Builders, Tasks 6.11 and 6.12 |
systre/mof5_net.cgd, systre/mof5_net.systre.txt | net and Systre output for MOF-5 | Topology, Task 5.6 |
systre/rutile_net.cgd, systre/rutile_net.systre.txt | net and Systre output for rutile | Topology, Task 5.6 |
zeopp_reference.json | Zeo++ porosity reference numbers | Porosity, Task 5.2 |
zeo_radii.cc | Zeo++ radii table | Porosity radii, Task 5.2.2 |
License status
The full RCSR snapshots are bundled in data/rcsr/ with the maintainer's authorisation (2026-11-09) and their checksums; the licensing note is in THIRD_PARTY.md. The Zeo++ radii table and the PORMAKE recording are not copied, and crates/xtalforge-mof/data/ contains hand-written illustrative blocks rather than the upstream package. The DFTB+ and EQeq files are reference outputs, not third-party data sets; their origin is recorded when they are copied.
Copy rule
A phase copies an item only when its tests use it. The copy lands with the subtask that introduces the test and names its source in the provenance file. The inventory is reviewed when a new phase starts.