tiled 0.1.0a8

Tile-based access to SciPy/PyData data structures over the web in many formats

Tiled

Disclaimer: This is very early work, still in the process of defining scope.

Data analysis is easier and better when we load and operate on data in simple, self-describing structures that keep our mind on the science rather the book-keeping of filenames and file formats.

Tiled is a data access tool that enables search and structured, chunkwise access to data in a variety of formats, regardless of the format the data happens to be stored in at rest. Like Jupyter, Tiled can be used solo or deployed as a shared resource. Tiled provides data, locally or over the web, in a choice of formats, spanning slow but widespread interchange formats (e.g. CSV, JSON, TIFF) and fast, efficient ones (e.g. C buffers, Apache Arrow DataFrames). Tiled enables slicing and sub-selection for accessing only the data of interest, and it enables parallelized download of many chunks at once. Users can access data with very light software dependencies and fast partial downloads.

Serve a "Catalog", a Python object backed by some generated data, directory of files, network resource, or database

tiled serve pyobject tiled.examples.generated:demo

And then access the data efficiently via the Python client, a web browser, or any HTTP client.

>>> from tiled.client import from_uri

>>> catalog = from_uri("http://localhost:8000")

>>> catalog
<Catalog {'arrays', 'dataframes', 'xarrays', 'nested', ...} ~5 entries>

>>> catalog['arrays']
<Catalog {'large', 'medium', 'small', 'tiny'}>

>>> catalog['arrays']['medium']
<ClientDaskArrayAdapter>

>>> catalog['arrays']['medium'][:]
array([[0.21267816, 0.59685753, 0.12483017, ..., 0.74891246, 0.43889019,
        0.27761903],
       [0.95434218, 0.31376234, 0.05776443, ..., 0.53886856, 0.92855426,
        0.32506382],
       [0.0458622 , 0.0561961 , 0.3893611 , ..., 0.23124064, 0.40311252,
        0.22488572],
       ...,
       [0.91990991, 0.98361972, 0.26394368, ..., 0.86427576, 0.00436757,
        0.03021872],
       [0.26595236, 0.18207517, 0.18989639, ..., 0.16221733, 0.59052007,
        0.94255651],
       [0.4721781 , 0.01424852, 0.57294198, ..., 0.70392867, 0.69371454,
        0.228491  ]])

>>> catalog['dataframes']
<Catalog {'df'}>

>>> catalog['dataframes']['df']
<ClientDaskDataFrameAdapter ['A', 'B', 'C']>

>>> catalog['dataframes']['df'][['A', 'B']]
              A         B
index                    
0      0.748885  0.769644
1      0.071319  0.364743
2      0.322665  0.897854
3      0.328785  0.810159
4      0.158253  0.822505
...         ...       ...
95     0.913758  0.488304
96     0.969652  0.287850
97     0.769774  0.941785
98     0.350033  0.052412
99     0.356245  0.683540

[100 rows x 2 columns]

Using an Internet browser or a commandline HTTP client like curl or httpie you can download the data in whole or in efficiently-chunked parts in the format of your choice:

# Download tabular data as CSV
http://localhost:8000/dataframe/full/dataframes/df?format=csv

# or XLSX (Excel)
http://localhost:8000/dataframe/full/dataframes/df?format=xslx

# and subselect columns.
http://localhost:8000/dataframe/full/dataframes/df?format=xslx&column=A&column=B

# View or download (2D) array data as PNG
http://localhost:8000/array/full/arrays/medium?format=png

# and slice regions of interest.
http://localhost:8000/array/full/arrays/medium?format=png&slice=:50,100:200

Web-based data access usually involves downloading complete files, in the manner of Globus; or using modern chunk-based storage formats, such as TileDB and Zarr in local or cloud storage; or using custom solutions tailored to a particular large dataset. Waiting for an entire file to download when only the first frame of an image stack or a certain column of a table are of interest is wasteful and can be prohibitive for large longitudinal analyses. Yet, it is not always practical to transcode the data into a chunk-friendly format or build a custom tile-based-access solution. (Though if you can do either of those things, you should consider them instead!)

In more technical language, Tiled is a service providing "tiled" chunk-based access to strided arrays, tablular datasets ("dataframes"), and nested structures thereof. It is centered on these structures (backed by numpy, pandas, xarray, various mappings in the server) rather than particular formats. The structures may be read from an extensible range of formats; the web client receives them as one of an extensible range of MIME types, and it can pose requests that sub-select and slice the data before it is read or served. The server incorporates search capability, which may be backed by a proper database solution at large scales or a simple in-memory index at small scales, as well as access controls. A Python client provides a friendly h5py-like interface and supports offline caching.