|
18 | 18 | - An "H →" reference is on screen wherever an arrow is. |
19 | 19 | - A project with no magnetic layer is unchanged: no arrows, no slider, and the |
20 | 20 | same Structure and SLD layout as before. |
| 21 | +- Multiple contrasts ([#425](https://github.com/easyscience/reflectometry-lib/issues/425)): |
| 22 | + - New demo files `examples/datasets/film_contrasts.ort` (a film on silicon in |
| 23 | + D2O, H2O and CMSi, with its slab model in the header) and `dppc_contrasts.ort` |
| 24 | + (three contrasts of a DPPC monolayer) for trying Add contrast and a joint fit |
| 25 | + (see the datasets README). |
| 26 | + - "Add contrast of the selected model…" on the Sample page creates a model |
| 27 | + sharing the selected one's structure, with chosen materials replaced (e.g. |
| 28 | + D2O by H2O, which also changes every hydrated layer) or re-formulated |
| 29 | + (isotopic substitution). Its scale, background and resolution are its own. |
| 30 | + A shared assembly is marked ⇄ in the model editor: editing it edits every |
| 31 | + model sharing it. |
| 32 | + - "Model constraints" is now "Link contrasts": it ties the selected models' |
| 33 | + corresponding parameters, but no longer their scale and background, and |
| 34 | + their material SLDs only on request. Layouts that do not correspond are |
| 35 | + reported instead of being tied partially. Removing a tie restores the |
| 36 | + parameter's value and bounds from before the link. |
| 37 | + - The parameters table lists a parameter shared by several models once, |
| 38 | + marked ⇄ with the models in its tooltip; a shared layer parameter can be |
| 39 | + detached for one model. |
| 40 | + - Each experiment's model is chosen per row on the Experiment and Analysis |
| 41 | + pages and shows the stored pairing (the combo used to reset to the first |
| 42 | + model and could change another experiment). |
| 43 | + - A "Fit" checkbox per experiment chooses what the next fit, or Bayesian |
| 44 | + sampling, includes; a parameter that an included model follows stays |
| 45 | + fitted even when its own model is left out. The pre-fit bound checks look |
| 46 | + only at the parameters the fit varies. The results dialog lists each |
| 47 | + dataset's χ² per point and share of the total; the statistics come from |
| 48 | + the library's record of the last completed run, so they survive switching |
| 49 | + the current model and are not replaced by a failed or cancelled run. |
| 50 | + - The Analysis chart has the staggered view too (not in the R·q⁴ view). |
| 51 | + - Linking several models links all of them or, when one cannot be linked, |
| 52 | + none. |
| 53 | + - Removing a model that experiments use asks whether to remove them or fit |
| 54 | + them with another model; experiments are no longer removed by position. |
| 55 | + - A data file that cannot be read is reported in a dialog. |
| 56 | + - The HTML report lists all models, with the model each contrast was derived |
| 57 | + from. |
| 58 | +- The legacy synchronous fit path (`Fitting.start_stop`) was removed; the GUI |
| 59 | + always fitted in a worker thread. |
21 | 60 |
|
22 | 61 | - Added a **Structure** tab on the Model page: a schematic view of the layer stack with one |
23 | 62 | colored box per layer (colors per material, heights following thickness, "× N" badges for |
|
0 commit comments