This is a series of two posts repeating some content I wrote up back in the Bioclipse days\n(see also this Scholia page). They both deal with something\nwe were facing: restructuring of version control repositories, while actually keeping the history. For\nexample, you may want to copy or move code from one repository to another. A second use case can be a file\nthat must be removed (there are valid reasons for that). Because these posts are based on Bioclipse work,\nthere will be some specific terminology, but the approach I regularly apply in other situations.
\n\nThis second post talks about how to migrate code from one repository to another.
\n\nWhile the above works pretty well, a good alternative in situations where you only need to get a\nrepository-with-history for a few plugins, is to use patch sets.
\n\nFirst, initialize a new git repository, e.g. bioclipse.rdf:
\n\nmkdir bioclipse.rdf\ncd bioclipse.rdf\ngit init\nnano README\ngit commit -m \"Added README with some basic info about the new repository\" README\nThen, for each plugin discover you need what the commit was where the plugins was first commited, using the git-svn repository created earlier:
\n\ncd your.gitsvn.checkout\ngit log --pretty=oneline externals/com.hp.hpl.jena/ | tail -1\nThen create patches for the last tree before that last patch by appending ^1 to the commit hash. For example, the first patch of the Jena libraries was 06d0eb0542377f958d06892860ea3363e3316389, so I type:
rm 00*.patch\ngit format-patch 06d0eb0542377f958d06892860ea3363e3316389^1 -- externals/com.hp.hpl.jena\n(tune the filter when removing old patches if there are more than 99!)
\n\nThe previous two steps can be combined into a Perl script:
\n\n#!/usr/bin/perl\nuse diagnostics;\nuse strict;\n\nmy $plugin = $ARGV[0];\n\nif (!$plugin) {\n print \"Syntax: gfp <plugin|feature>\\n\";\n exit(0);\n}\n\ndie \"Cannot find plugin or feature $plugin !\" if (!(-e $plugin));\n\n`rm -f *.patch`;\nmy $hash = `git log --follow --pretty=oneline $plugin | tail -1 | cut -d' ' -f1`;\n$hash =~ s/\\n|\\r//g;\n\nprint \"Plugin: $plugin \\n\";\nprint \"Hash: $hash \\n\";\n`git format-patch $hash^1 -- $plugin`;\nMove these patches into your new repository:
\n\nmv 00*.patch ../bioclipse.rdf\n(tune the filter when moving the patches if there are more than 99! Also customize the target folder name to match your situation)
\n\nApply the new patches in your new git repository:
\n\ncd ../bioclipse.rdf\ngit am 00*.patch\n(You’re on your own if that fails… and you may have to default to the other alternative then)
\n\nRepeat those two steps for all plugins you want in your new repository
", "summary": "This is a series of two posts repeating some content I wrote up back in the Bioclipse days (see also this Scholia page). They both deal with something we were facing: restructuring of version control repositories, while actually keeping the history. For example, you may want to copy or move code from one repository to another. A second use case can be a file that must be removed (there are valid reasons for that). Because these posts are based on Bioclipse work, there will be some specific terminology, but the approach I regularly apply in other situations.", "date_published": "2020-07-03T00:00:00+00:00", "date_modified": "2025-02-15T00:00:00+00:00", "tags": ["bioclipse","git"], "authors": [ { "name": "Egon Willighagen", "url": "https://orcid.org/0000-0001-7542-0286" } ] } ] }