shed_diff smart diff not hiding dependency changes (changeset_revision etc) #205
Description
The help suggests smart/magic diff is the default,
$ planemo shed_diff --help
...
--raw Do not attempt smart diff of XML to filter out
attributes populated by the Tool Shed.
...
However, this isn't being smart enough:
$ planemo shed_diff --shed_target testtoolshed ~/repositories/pico_galaxy/tools/sample_seqs/
wget -q --recursive -O - 'https://testtoolshed.g2.bx.psu.edu/repository/download?repository_id=de64f62614881770&changeset_revision=default&file_type=gz' | tar -xzf - -C /tmp/tool_shed_diff_i5xHt_/_testtoolshed_ --strip-components 1
mkdir "/tmp/tool_shed_diff_i5xHt_/_local_"; tar -xzf "/tmp/tmpS0q0pC" -C "/tmp/tool_shed_diff_i5xHt_/_local_"; rm -rf /tmp/tmpS0q0pC
cd "/tmp/tool_shed_diff_i5xHt_"; diff -r _local_ _testtoolshed_
diff -r _local_/tools/sample_seqs/tool_dependencies.xml _testtoolshed_/tools/sample_seqs/tool_dependencies.xml
4c4
< <repository name="package_biopython_1_65" owner="biopython" />
---
> <repository changeset_revision="f8d72690eeae" name="package_biopython_1_65" owner="biopython" toolshed="https://testtoolshed.g2.bx.psu.edu" />
This looks like the raw diff output:
$ planemo shed_diff --shed_target testtoolshed ~/repositories/pico_galaxy/tools/sample_seqs/ --raw
wget -q --recursive -O - 'https://testtoolshed.g2.bx.psu.edu/repository/download?repository_id=de64f62614881770&changeset_revision=default&file_type=gz' | tar -xzf - -C /tmp/tool_shed_diff_VGHJk0/_testtoolshed_ --strip-components 1
mkdir "/tmp/tool_shed_diff_VGHJk0/_local_"; tar -xzf "/tmp/tmpBmc0Ck" -C "/tmp/tool_shed_diff_VGHJk0/_local_"; rm -rf /tmp/tmpBmc0Ck
cd "/tmp/tool_shed_diff_VGHJk0"; diff -r _local_ _testtoolshed_
diff -r _local_/tools/sample_seqs/tool_dependencies.xml _testtoolshed_/tools/sample_seqs/tool_dependencies.xml
4c4
< <repository name="package_biopython_1_65" owner="biopython" />
---
> <repository changeset_revision="f8d72690eeae" name="package_biopython_1_65" owner="biopython" toolshed="https://testtoolshed.g2.bx.psu.edu" />
I think the problem is the assumption in planemo/shed/diff.py that the special files tool_dependencies.xml and repository_dependencies.xml will only be at the tar-ball root:
def diff_and_remove(working, label_a, label_b, f):
a_deps = os.path.join(working, label_a, "tool_dependencies.xml")
b_deps = os.path.join(working, label_b, "tool_dependencies.xml")
a_repos = os.path.join(working, label_a, "repository_dependencies.xml")
b_repos = os.path.join(working, label_b, "repository_dependencies.xml")
Also, as a possible enhancement, should we consider the TS doing attribute reordering, which prompted me to make this change (testing with an older version of planemo)?: peterjc/pico_galaxy@698859a