Serves one or more XML/HDF5 datasets for remote access over HTTP. Provide (NAME XML) pairs on the command line or in a dataset file,
where NAME is the name under which the dataset should be made accessible and XML is the path to the XML file of the dataset.
### Command line options
-`-b <BASEURL>` - Base URL under which the server will be made visible (e.g., if behind a proxy).
-`-d <FILE>` - Dataset file: A plain text file specifying one dataset per line. Each line is formatted as "NAME `<TAB>` XML".
-`-p <PORT>` - Listening port (default: 8080).
-`-s <HOSTNAME>` - Hostname of the server (default: `<COMPUTER-NAME>`).
-`-t <DIRECTORY>` - Directory to store thumbnails (new temporary directory by default.).
## QCMP Compression Extension
The QCMP compression extension of BigDataViewer Web Server gives user an option to use lossy compression, which greatly reduce the amount of data transfered through the network. Vector quantization with Huffman coding is used to compress the data.
In order to be able to use compression for your dataset, the dataset codebook for given quantization must be prepared. The codebook can be created with the QCMP console application using the `-tcb,--train-codebook` CLI method. For further details see [compression library source code and README](https://code.it4i.cz/BioinformaticDataCompression/QcmpCompressionLibrary). The codebooks are required to be created before hand, because their creation is quite time consuming.
### Additional command line options
-`-qcmp` - Enables the QCMP compression on the server.
-`-cbc, --codebook-cache <DIRECTORY>` - Directory with prepared codebooks.
-`-vq, --vector-quantization <VECTOR>` - Set vector quantization as the used lossy compression algorithm. *REQUIRED FOR NOW*. `VECTOR` specifies the dimenions of the quantization vector eg. 3x3, 3x3x3, ...
-`-v, --verbose` - Make the QCMP cell handler verbose.
-`-cf,--compress-from <PYRAMID-LEVEL>` - Set minimal mipmap/pyramid level, which should be compressed. For example `-cf 1` means that level 0 won't be compressed, but levels greater and equal to 1 will be.
### Example:
We have dataset *drosophila32* with its XML file drosophila_32x32.xml and we want to use vector quantization with quantization vector of dimensions 3x3x3. This is three-dimensional *voxel* vector with 27 values. We have exported our dataset using BigDataViewer's `Export Current Image as XML/HDF5` tool. We recommend to check `manual mipmap setup` and set bigger Hdf5 chunk sizes, for example we have used {32,32,32} for all mipmap levels. Bigger chunks would normally result in greater data transfer (8 KB for $16^3$ and 65 KB for $32^3$), but compression greatly reduces the chunk size. Also we recommend to create as many codebooks as there are subsampling factors or pyramid levels. By multiple codebooks we mean codebooks of different size. Larger codebook means better compression. Possible codebook sizes are (*for now*) 256, 128, 64, 32, 16, 8 and 4. If we export our dataset with four pyramid levels/subsampling factors, four largest codebooks would be used 256, 128, 64 and 32.
Subsampling factor / pyramid level mapping to codebook size example:
```
Level 0 --> Codebook size 256 (Best quality)
Level 1 --> Codebook size 128
Level 2 --> Codebook size 64
Level 3 --> Codebook size 32
```
If we choose to set `--compress-from 1` the mapping changes accordingly:
```
Level 0 --> Uncompressed data
Level 1 --> Codebook size 256
Level 2 --> Codebook size 128
Level 3 --> Codebook size 64
```
To start a local server with enabled QCMP compression we would then use these options:
