diff options
Diffstat (limited to 'scripts/pd-diff')
-rwxr-xr-x | scripts/pd-diff | 44 |
1 files changed, 44 insertions, 0 deletions
diff --git a/scripts/pd-diff b/scripts/pd-diff index 26db0a18..a5d6fa22 100755 --- a/scripts/pd-diff +++ b/scripts/pd-diff @@ -1,5 +1,49 @@ #!/bin/sh +# difference between two Pd patches + +# the problem here is, that there exists a many-to-1 relation between +# Pd-file and patch + +# e.g. objects might have a different creation order without affecting +# the functionality of a patch +# otoh, creation order might have implications as well +# nevertheless, creation order has an impact on the Pd-file, since connections +# between objects are made using indices; therefore a change in the creation order +# of the objects will be reflected on several places within the Pd-file! + +# furthermore, in most cases the position of an object within the patch does not +# have any significance +# the noteable built-in exceptions are [inlet] and [outlet] + + + +# these weirdnesses make it hard to effectively find out whether a patch description +# differs substantially from another one +# in practice, traditional diff will only tell you that _something_ has changed, +# but it is hard to interprete its output in a meaningful way + +# even harder it is to build a merger of Pd-files (for concurrent development) + + +## originally draft for this differ (as i deduce it from hcs's code) +# +# 2 pass evaluation: +# - find diffs in the object-list without position information +# - count connections in each patch and compare them + + +# TODO +# - more algorithms for diffing features +# - modularity (allow user to chose diff-algorithms) +# - better tempfile algorithm (ideally there wouldn't be any tempfiles at all) +# - cleanup at the end (no tempfiles should be left behind) +# - ensure that this script is portable +# (probably switch from "shell" (bash) to another language, like tcl + + + + DATE=$(date '+%Y-%m-%d_%H.%M.%S') TMPDIR=/tmp/pd-diff |