Make datatype H5_Model for machine leaning models

lib/galaxy/config/sample/datatypes_conf.xml.sample

@@ -170,6 +170,7 @@
     <datatype extension="hivtrace" type="galaxy.datatypes.text:Json" mimetype="application/json" subclass="true" display_in_upload="false"/>
     <datatype extension="cool" type="galaxy.datatypes.binary:Cool" mimetype="application/octet-stream" display_in_upload="true"/>
     <datatype extension="mcool" type="galaxy.datatypes.binary:MCool" mimetype="application/octet-stream" display_in_upload="true"/>
+    <datatype extension="h5_model" type="galaxy.datatypes.binary:H5_Model" mimetype="application/octet-stream" display_in_upload="true"/>
     <datatype extension="html" type="galaxy.datatypes.text:Html" mimetype="text/html"/>
     <datatype extension="interval" type="galaxy.datatypes.interval:Interval" display_in_upload="true" description="File must start with definition line in the following format (columns may be in any order).">
       <converter file="interval_to_bed_converter.xml" target_datatype="bed"/>
@@ -907,6 +908,7 @@
     <sniffer type="galaxy.datatypes.binary:NcbiTaxonomySQlite"/>
     <sniffer type="galaxy.datatypes.binary:SQlite"/>
     <sniffer type="galaxy.datatypes.binary:Cool"/>
+    <sniffer type="galaxy.datatypes.binary:H5_Model"/>
     <sniffer type="galaxy.datatypes.binary:MCool"/>
     <sniffer type="galaxy.datatypes.binary:Loom"/>
     <sniffer type="galaxy.datatypes.binary:Anndata"/>

lib/galaxy/datatypes/binary.py

@@ -3,6 +3,7 @@
 import binascii
 import gzip
 import io
+import json
 import logging
 import os
 import shutil
@@ -26,7 +27,13 @@ from galaxy.datatypes.data import (
     DatatypeValidation,
     get_file_peek,
 )
-from galaxy.datatypes.metadata import DictParameter, ListParameter, MetadataElement, MetadataParameter
+from galaxy.datatypes.metadata import (
+    DictParameter,
+    FileParameter,
+    ListParameter,
+    MetadataElement,
+    MetadataParameter,
+)
 from galaxy.datatypes.sniff import build_sniff_from_prefix
 from galaxy.util import nice_size, sqlite
 from galaxy.util.checkers import is_bz2, is_gzip
@@ -1452,6 +1459,71 @@ class MCool(H5):
             return "MCool (HDF5) file (%s)." % (nice_size(dataset.get_size()))
 
 
+class H5_Model(H5):
+    """
+    Machine learning models generated by Galaxy-ML.
+    """
+    file_ext = "h5_model"
+
+    MetadataElement(name="params", desc="Hyperparameter File", param=FileParameter, file_ext="tabular", readonly=True, no_value=None, visible=False, optional=True)
+
+    def set_meta(self, dataset, overwrite=True, **kwd):
+        try:
+            spec_key = "params"
+            params_file = dataset.metadata.params
+            if not params_file:
+                params_file = dataset.metadata.spec[spec_key].param.new_file(dataset=dataset)
+            with h5py.File(dataset.file_name, "r") as handle:
+                params = handle["-model_hyperparameters-"][()]
+            params = json.loads(params.decode('utf-8'))
+            with open(params_file.file_name, "w") as f:
+                f.write("\tParameter\tValue\n")
+                for p in params:
+                    f.write("\t".json(p) + "\n")
+            dataset.metadata.params_file = params_file
+        except Exception as e:
+            log.warning('%s, set_meta Exception: %s', self, e)
+
+    def sniff(self, filename):
+        URL = "https://github.com/goeckslab/Galaxy-ML"
+
+        if super().sniff(filename):
+            keys = ["-model_config-"]
+            with h5py.File(filename, "r") as handle:
+                if not all(name in handle.keys() for name in keys):
+                    return False
+                url = util.unicodify(handle.attrs.get("-URL-"))
+            if url == URL:
+                return True
+        return False
+
+    def get_config(self, filename):
+        try:
+            with h5py.File(filename, "r") as handle:
+                config = util.unicodify(handle["-model_config-"][()])
+            return config
+        except Exception as e:
+            log.warning('%s, get model configuration Except: %s', self, e)
+            return ""
+
+    def set_peek(self, dataset, is_multi_byte=False):
+        if not dataset.dataset.purged:
+            config = self.get_config(dataset.file_name)
+            if len(config) > 100:
+                config = config[:100]
+            dataset.peek = config
+            dataset.blurb = nice_size(dataset.get_size())
+        else:
+            dataset.peek = 'file does not exist'
+            dataset.blurb = 'file purged from disk'
+
+    def display_peek(self, dataset):
+        try:
+            return dataset.peek
+        except Exception:
+            return "HDF5 Model (%s)" % (nice_size(dataset.get_size()))
+
+
 class Scf(Binary):
     """Class describing an scf binary sequence file"""
     edam_format = "format_1632"