Would this tool be useful - encoding repository data into single URL

Sebastian Gniazdowski· May 9, 2017, 12:48 UTC · lore

Hello I wonder about usability of following tool. Quick-start:

    giturl https://github.com/zdharma/giturl -r devel -p lib/coding_functions.cpp

    Protocol: https
    Site:     github.com
    Repo:     zdharma/giturl
    Revision: devel
    File:     lib/coding_functions.cpp

    gitu://ŬṽǚǫoŒẗ6ẏȅcЭÑẩőn4ầŘїệαЃȣϟṈӛŀї

It does Huffman encoding and base-1024 encoding to pack given data into single URL. The Unicode characters selected for base-1024 encoding are letters, not symbols, so double-clicking in e.g. web browser selects the whole code, making it easy to grab a repository data.

Decoding:

    giturl -qd ŬṽǚǫoŒẗ6ẏȅcЭÑẩőn4ầŘїệαЃȣϟṈӛŀї
    https://github.com/zdharma/giturl / rev:devel / file:lib/coding_functions.cpp

I can also encode commits relative to given revision, e.g. bits 10011 are commits 1, 4, 5. Easy to add to the g-code. Selecting 10th commit is only 1 character in base-1024.

However I wonder if this has any uses. Could be patches sent this way? Having refs/patches/<name>, encoding <name> in URL, sending it instead of inlining/attaching a diff, selecting e.g. 3 commits via the bit-mask mentioned. That said, it's more about easy-grab of repository data and storage in well-defined, consistent format, not in language "the branch is ..., commit a7a35cb". Does this make sense?

There are 2 implementations, in Zsh (uses Zshell like e.g. Ruby, not interactively) and C++11 (mostly because of std::regex):

https://github.com/zdharma/giturl https://github.com/zdharma/cgiturl

-- Sebastian Gniazdowski psprint /at/ zdharma.org

Ævar Arnfjörð Bjarmason· May 9, 2017, 13:40 UTC · re: Sebastian Gniazdowski · lore

Re: Would this tool be useful - encoding repository data into single URL

On Tue, May 9, 2017 at 2:48 PM, Sebastian Gniazdowski <psprint@zdharma.org> wrote:

Show 28 quoted lines

> Hello
> I wonder about usability of following tool. Quick-start:
>
>     giturl https://github.com/zdharma/giturl -r devel -p lib/coding_functions.cpp
>
>     Protocol: https
>     Site:     github.com
>     Repo:     zdharma/giturl
>     Revision: devel
>     File:     lib/coding_functions.cpp
>
>     gitu://ŬṽǚǫoŒẗ6ẏȅcЭÑẩőn4ầŘїệαЃȣϟṈӛŀї
>
> It does Huffman encoding and base-1024 encoding to pack given data into single URL. The Unicode characters selected for base-1024 encoding are letters, not symbols, so double-clicking in e.g. web browser selects the whole code, making it easy to grab a repository data.
>
> Decoding:
>
>     giturl -qd ŬṽǚǫoŒẗ6ẏȅcЭÑẩőn4ầŘїệαЃȣϟṈӛŀї
>     https://github.com/zdharma/giturl / rev:devel / file:lib/coding_functions.cpp
>
> I can also encode commits relative to given revision, e.g. bits 10011 are commits 1, 4, 5. Easy to add to the g-code. Selecting 10th commit is only 1 character in base-1024.
>
> However I wonder if this has any uses. Could be patches sent this way? Having refs/patches/<name>, encoding <name> in URL, sending it instead of inlining/attaching a diff, selecting e.g. 3 commits via the bit-mask mentioned. That said, it's more about easy-grab of repository data and storage in well-defined, consistent format, not in language "the branch is ..., commit a7a35cb". Does this make sense?
>
> There are 2 implementations, in Zsh (uses Zshell like e.g. Ruby, not interactively) and C++11 (mostly because of std::regex):
>
> https://github.com/zdharma/giturl
> https://github.com/zdharma/cgiturl

This looks really cool. Yeah patches can be sent this way, see Documentation/SubmittingPatches in the git.git repository, online version: https://github.com/git/git/blob/master/Documentation/SubmittingPatches