Merge pull request #18080 from martenson/net-data

add Net datatype
galaxyproject · May 3, 2024 · 72e00b2 · 72e00b2
2 parents 65559a8 + d4aa9c4
commit 72e00b2
Show file tree

Hide file tree

Showing 3 changed files with 96 additions and 4 deletions.
diff --git a/lib/galaxy/config/sample/datatypes_conf.xml.sample b/lib/galaxy/config/sample/datatypes_conf.xml.sample
@@ -988,7 +988,8 @@
     <datatype extension="shp" type="galaxy.datatypes.gis:Shapefile" mimetype="application/octet-stream" display_in_upload="true" description="geospatial vector data format for geographic information system"/>
     <!-- Flexible Image Transport System (FITS) used in Astronomy https://fits.gsfc.nasa.gov/ https://fits.gsfc.nasa.gov/rfc4047.txt -->
     <datatype extension="fits" type="galaxy.datatypes.binary:FITS" mimetype="application/octet-stream" display_in_upload="true" description="Flexible Image Transport System (FITS) used in Astronomy"/>
-    <datatype extension="chain" type="galaxy.datatypes.chain:Chain" display_in_upload="true"/>
+    <datatype extension="chain" type="galaxy.datatypes.chain:Chain" display_in_upload="true" description_url="https://genome.ucsc.edu/goldenPath/help/chain.html"/>
+    <datatype extension="ucsc.net" type="galaxy.datatypes.chain:Net" display_in_upload="true" description_url="https://genome.ucsc.edu/goldenPath/help/net.html"/>
   </registration>
   <sniffers>
     <!--
@@ -1084,7 +1085,7 @@
     <sniffer type="galaxy.datatypes.binary:Edr"/>
     <sniffer type="galaxy.datatypes.binary:Vel"/>
     <sniffer type="galaxy.datatypes.binary:Xlsx"/>
-    <sniffer type="galaxy.datatypes.binary:Numpy"/>    
+    <sniffer type="galaxy.datatypes.binary:Numpy"/>
     <sniffer type="galaxy.datatypes.qiime2:QIIME2Metadata"/>
     <sniffer type="galaxy.datatypes.qiime2:QIIME2Artifact"/>
     <sniffer type="galaxy.datatypes.qiime2:QIIME2Visualization"/>
@@ -1093,6 +1094,7 @@
     <sniffer type="galaxy.datatypes.annotation:Augustus"/>
     <sniffer type="galaxy.datatypes.xml:Owl"/>
     <sniffer type="galaxy.datatypes.chain:Chain"/>
+    <sniffer type="galaxy.datatypes.chain:Net"/>
     <sniffer type="galaxy.datatypes.triples:Rdf"/>
     <sniffer type="galaxy.datatypes.blast:BlastXml"/>
     <sniffer type="galaxy.datatypes.images:Gifti" />

diff --git a/lib/galaxy/datatypes/chain.py b/lib/galaxy/datatypes/chain.py
@@ -1,5 +1,5 @@
 """
-Genome browser chain format class
+Genome browser alignment formats
 """
 
 import logging
@@ -98,7 +98,6 @@ def sniff_prefix(self, file_prefix: FilePrefix) -> bool:
             line = line.strip()
             if line:  # first non-empty line
                 if line.startswith("chain"):
-                    # The next line.strip() must not be '', nor startwith '>'
                     tokens = line.split()
                     if not (
                         len(tokens) in [12, 13]
@@ -127,3 +126,69 @@ def sniff_prefix(self, file_prefix: FilePrefix) -> bool:
                 else:
                     return False
         return False
+
+
+@build_sniff_from_prefix
+class Net(data.Text):
+    """Class describing a net format alignment file"""
+
+    edam_format = "format_3983"
+    file_ext = "ucsc.net"
+
+    def sniff_prefix(self, file_prefix: FilePrefix) -> bool:
+        """
+        Determines whether the file is in net format
+
+        For details see https://genome.ucsc.edu/goldenPath/help/net.html
+
+        Rules for sniffing as True:
+
+            We don't care about line length (other than empty lines).
+
+            The first non-empty line must start with 'net' followed by chromName (str) and chromSize (int)
+
+            We will only check that the first "net" line and the first data line are formatted correctly.
+
+        >>> from galaxy.datatypes.sniff import get_test_fname
+        >>> fname = get_test_fname( '1.chain' )
+        >>> Net().sniff( fname )
+        False
+        >>> fname = get_test_fname( '1.ucsc.net' )
+        >>> Net().sniff( fname )
+        True
+        >>>
+        """
+        allowed_classes = ["fill", "gap"]
+        strands = ["+", "-"]
+
+        fh = file_prefix.string_io()
+        for line in fh:
+            line = line.strip()
+            if line:  # first non-empty line
+                if line.startswith("net"):
+                    tokens = line.split()
+                    if not (len(tokens) == 3 and tokens[2].isdigit()):
+                        return False
+                    for line in fh:
+                        if line[0] != " ":  # children are indented one space
+                            return False
+                        line = line.strip()
+                        if line == "":
+                            break
+                        tokens = line.split()
+                        if not (
+                            len(tokens) >= 7  # seven fixed fields
+                            and len(tokens) <= 41  # plus seventeen optional name/value pairs
+                            and tokens[0] in allowed_classes
+                            and tokens[1].isdigit()
+                            and tokens[2].isdigit()
+                            and tokens[4] in strands
+                            and tokens[5].isdigit()
+                            and tokens[6].isdigit()
+                        ):
+                            return False
+                        else:
+                            return True
+                else:
+                    return False
+        return False
diff --git a/lib/galaxy/datatypes/test/1.ucsc.net b/lib/galaxy/datatypes/test/1.ucsc.net
@@ -0,0 +1,25 @@
+net chr2L 23011544
+ fill 6004 3278 chrXR_group3a - 1396397 2164 id 25606 score 23114 ali 782 qDup 576 type top tN 0 qN 0 tR 36 qR 0 tTrf 0 qTrf 0
+  gap 6065 2 chrXR_group3a - 1398498 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+  gap 6096 1485 chrXR_group3a - 1397572 897 tN 0 qN 0 tR 36 qR 0 tTrf 0 qTrf 0
+   fill 6096 513 chrU - 5570675 533 id 48675 score 4435 ali 465 qDup 533 type nonSyn tN 0 qN 0 tR 0 qR 13 tTrf 0 qTrf 0
+    gap 6116 8 chrU - 5571188 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 6156 5 chrU - 5571156 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 6184 3 chrU - 5571133 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 6212 18 chrU - 5571106 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 6244 9 chrU - 5571092 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 6340 2 chrU - 5570996 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 6515 3 chrU - 5570771 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+  gap 7623 1 chrXR_group3a - 1397530 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+  gap 7664 1007 chrXR_group3a - 1397008 482 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+   fill 7664 382 chrXL_group1e - 8262003 506 id 25608 score 10609 ali 364 qDup 506 type nonSyn tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 7784 4 chrXL_group1e - 8262361 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 7792 3 chrXL_group1e - 8262357 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 7921 2 chrXL_group1e - 8262126 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+    gap 7949 9 chrXL_group1e - 8262092 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+  gap 8693 1 chrXR_group3a - 1396985 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+ fill 9833 1251 chrU - 5562980 1239 id 48675 score 10720 ali 1124 qDup 1094 type top tN 0 qN 0 tR 22 qR 88 tTrf 0 qTrf 0
+  gap 9966 7 chrU - 5564075 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+  gap 10015 3 chrU - 5564030 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+  gap 10088 2 chrU - 5563957 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0
+  gap 10101 8 chrU - 5563946 0 tN 0 qN 0 tR 0 qR 0 tTrf 0 qTrf 0