# Data Structure

Ewokstomo tasks do not rely on a specific file structure. The entry point
of an ewokstomo workflow is either an HDF5 file (for insitute specific tasks), or
more commonly an .nx file (NXtomo). From this path, the workflow will save the results
in specific folder following this structure:

``` text
.
├── parent/
│   ├── <dataset>.nx              # NXtomo file produced by H5ToNx
│   └── gallery/                  # Projection thumbnails (200x200 + *_large) + projections GIF
├── references/
│   ├── <dataset>_darks.hdf5      # Reduced darks from ReduceDarkFlat
│   └── <dataset>_flats.hdf5      # Reduced flats from ReduceDarkFlat
└── slices/
    ├── <prefix>/<prefix>_<idx>.hdf5  # Single-slice reconstruction from ReconstructSlice
    └── gallery/                      # Slice thumbnails (200x200 + *_large)
```

Several things can be noted: 
- `<prefix>` is built from the NX filename and technique
  (e.g. `TestEwoksTomo_0010_absorption_xy_00008`).
- Gallery tasks always emit a full-size `_large` image plus a 200x200
  resized version.
- Parents are created automatically; you can point `nx_path` or
  `process_folder_path` to any writable location that fits the structure
  above.

Here is where here each tasks writes:

- (ESRF-only) `H5ToNx` creates `projections/<dataset>.nx`.
- `ReduceDarkFlat` writes reduced references into
  `references/<dataset>_{darks,flats}.hdf5`.
- `BuildProjectionsGallery` stores projection thumbnails under
  `projections/gallery/` and writes `<dataset>_projections.gif` (200x200, 32 colors).
- `ReconstructSlice` writes a single slice under `slices/` and sets the
  Nabu output directory to that folder.
- `BuildSlicesGallery` places slice thumbnails under `slices/gallery/`.
- (ESRF-only) `DataPortalUpload` accepts either the processed dataset root or one of
  its `projections/` or `slices/` subfolders; it extracts metadata
  (beamline, proposal, sample, dataset) from the path when using the
  ESRF scheme.

````{note}

## ESRF-specific paths:

At ESRF the file structure is specific. The raw data (from the acquisition) is stored as:


```text
RAW_DATA/<sample>/<dataset>/
├── <dataset>.h5                  # Main acquisition (e.g. Bliss HDF5)
└── scan000X/*.h5                 # Folders containing lima files
```

Acquisition files live under `RAW_DATA/<sample>/<dataset>/` while workflow 
outputs live under `PROCESSED_DATA/<sample>/<dataset>/`.

The processed dataset mirrors the raw dataset name and groups the
artifacts produced by `ewokstomo` tasks:

```text
PROCESSED_DATA/<sample>/<dataset>/
├── projections/
│   ├── <dataset>.nx
│   └── gallery/                  # Thumbnails + <dataset>_projections.gif
├── references/
│   ├── <dataset>_darks.hdf5
│   └── <dataset>_flats.hdf5
└── slices/
    ├── <prefix>/<prefix>_<idx>.hdf5
    └── gallery/
```


````