Tile-based access to SciPy/PyData data structures over the web in many formats
Project description
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 common, self-describing structures that keep our mind on the science rather than the book-keeping of filenames and file formats.
Tiled is a data access service for data-aware portals and data science tools.
Tiled has a Python client and integrates naturally with dask and Python data science
libraries, but nothing about the service is Python-specific; it also works from
a web browser, curl, or any HTTP client.
Tiled’s service can sit atop databases, filesystems, and/or remote services to enable search and structured, chunkwise access to data in an extensible variety of appropriate formats, providing data in a consistent structure regardless of the format the data happens to be stored in at rest. The natively-supported formats span slow but widespread interchange formats (e.g. CSV, JSON) and fast, efficient ones (e.g. C buffers, Apache Arrow DataFrames). Tiled enables slicing and sub-selection to read and transfer 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.
Tiled puts an emphasis on structures rather than formats, including:
- N-dimensional strided arrays (i.e. numpy-like arrays)
- Tabular data (i.e. pandas-like "dataframes")
- Hierarchical structures thereof (e.g. xarrays, HDF5-compatible structures like NeXus)
Tiled implements extensible access control enforcement based on web security standards, similar to JuptyerHub. Like Jupyter, Tiled can be used by a single user or deployed as a shared public or private resource.
Tiled facilitates client-side caching in a standard web browser or in Tiled's Python client, making efficient use of bandwidth and enabling an offline "airplane mode." Service-side caching of "hot" datasets and resources is also possible.
| Distribution | Where to get it |
|---|---|
| PyPI | pip install tiled |
| Conda | Coming Soon |
| Source code | github.com/bluesky/tiled |
| Documentation | blueskyproject.io/tiled |
Example
In this example, we'll serve of a collection of data that is generated in memory. Alternatively, it could be read on demand from a directory of files, network resource, database, or some combination of these.
tiled serve pyobject --public 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
>>> client = from_uri("http://localhost:8000")
>>> client
<Node {'arrays', 'dataframes', 'xarrays', 'nested', ...} ~5 entries>
>>> client['arrays']
<Node {'large', 'medium', 'small', 'tiny'}>
>>> client['arrays']['medium']
<ArrayClient>
>>> client['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 ]])
>>> client['dataframes']
<Node {'df'}>
>>> client['dataframes']['df']
<DataFrameClient ['A', 'B', 'C']>
>>> client['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 command-line 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!)
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file tiled-0.1.0a29.tar.gz.
File metadata
- Download URL: tiled-0.1.0a29.tar.gz
- Upload date:
- Size: 128.4 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/3.4.2 importlib_metadata/4.6.0 pkginfo/1.5.0.1 requests/2.24.0 requests-toolbelt/0.9.1 tqdm/4.50.0 CPython/3.8.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 | e0146bd41d19f9cbf8d1a11c91efb1d74b4a1bba77059cdb1a6dbd5e0ba4ac08 |
|
| MD5 | b33a3e447f70eaf0051b2294a4655356 |
|
| BLAKE2b-256 | c3973c4db126ba480e6f8cd0ab3ee412c05ea063f30bddc3dd2db0fedbd2e100 |
Provenance
File details
Details for the file tiled-0.1.0a29-py3-none-any.whl.
File metadata
- Download URL: tiled-0.1.0a29-py3-none-any.whl
- Upload date:
- Size: 126.9 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/3.4.2 importlib_metadata/4.6.0 pkginfo/1.5.0.1 requests/2.24.0 requests-toolbelt/0.9.1 tqdm/4.50.0 CPython/3.8.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 | ca5e1ae91c78ac71e08291fb41f2cdc381d72dcbd679d8f11e2a806549c9780c |
|
| MD5 | fbb5d615083074c69d90a4cd636f604c |
|
| BLAKE2b-256 | d75ab44a0f6e1288edf4b6f15b10c43543e9d0505407c0444fe525c78d16a2f9 |
