From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.2 (2018-09-13) on dcvr.yhbt.net X-Spam-Level: X-Spam-ASN: AS53758 23.128.96.0/24 X-Spam-Status: No, score=-4.0 required=3.0 tests=AWL,BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS, MAILING_LIST_MULTI,SPF_HELO_PASS,SPF_PASS shortcircuit=no autolearn=ham autolearn_force=no version=3.4.2 Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by dcvr.yhbt.net (Postfix) with ESMTP id 8CE0A1F953 for ; Tue, 30 Nov 2021 20:43:48 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S244417AbhK3UrC (ORCPT ); Tue, 30 Nov 2021 15:47:02 -0500 Received: from out3-smtp.messagingengine.com ([66.111.4.27]:33841 "EHLO out3-smtp.messagingengine.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S233340AbhK3Uq7 (ORCPT ); Tue, 30 Nov 2021 15:46:59 -0500 Received: from compute5.internal (compute5.nyi.internal [10.202.2.45]) by mailout.nyi.internal (Postfix) with ESMTP id C471D5C0275; Tue, 30 Nov 2021 15:43:38 -0500 (EST) Received: from mailfrontend2 ([10.202.2.163]) by compute5.internal (MEProxy); Tue, 30 Nov 2021 15:43:38 -0500 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=u92.eu; h=from :to:cc:subject:date:message-id:mime-version :content-transfer-encoding; s=fm3; bh=2aECx+8T6niNOCoEietSP3/6BZ oLqwnPChMqKqZ8ye0=; b=l6/9BlV27TFCI4tXmyz6rQtDx0NOae5stK+EPCNP8H juUskb7Q0MsqePoj9v/xVQ6GCHX7PtpKynckJJifD8xB+KId+gC4BVMlKepA7PNH ouDsIgAZeOtSm/47EErzDRo7Leplydx2ViZFaFyEixq1LsVwn9W5XLt5fBnS+5TZ VOCucAo2R1YwdEG4uiMXolYYZKq7ukj/32/CiehLtqTcOnTyk2WOpQZw9xjz2bSA DRBobZxHf3izdLm540dxSvqiEwrCvrT3nYN6Dr0cRxKsCwNgBVIOuPkjjYQzOUj9 wMQ01nEbI7S0scVEseV/+gMGSFy/XTiUL8uZxaxTCInA== DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=cc:content-transfer-encoding:date:from :message-id:mime-version:subject:to:x-me-proxy:x-me-proxy :x-me-sender:x-me-sender:x-sasl-enc; s=fm1; bh=2aECx+8T6niNOCoEi etSP3/6BZoLqwnPChMqKqZ8ye0=; b=Z6KLf9AAX7DPCwdDtmVnee+Edw6U3HeTo b24YepvOfuRyH3j8A161eJbnOu7rMz5kb8IB7+t7ndw3NUGzB17C5XpoNog1t1fo qstIpj2Gmg2bh0+I+gcljlQWzWgv9vgASZ20z1l6UMIFCL/4/+sQyZ7zevEOqCJh miRm1W+HmN4RGL3qt8K+JyanGxf75+STX7aswr1wrXCvBaEldil7TgiTM7wRo97Y 9uIMkZBzGGfWPv/k0mkyD7xjByK8wDeLc/joUNUD48foq9eiwM2XW09bBTELOboj Ji5iq23zjsaA1DL6jFYTtOFitg5luy0Vrl5s0xyELPUo15GuC/CMQ== X-ME-Sender: X-ME-Received: X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvuddriedugddugeduucetufdoteggodetrfdotf fvucfrrhhofhhilhgvmecuhfgrshhtofgrihhlpdfqfgfvpdfurfetoffkrfgpnffqhgen uceurghilhhouhhtmecufedttdenucenucfjughrpefhvffufffkofgggfestdekredtre dttdenucfhrhhomhephfgvrhhnrghnughoucftrghmohhsuceoghhrvggvnhhfohhosehu ledvrdgvuheqnecuggftrfgrthhtvghrnhepgfeifeduheegudehfeevfffhffelieefte effeehgeeileekuefhgeegueethfeinecuffhomhgrihhnpehkvghrnhgvlhdrohhrghdp phgrshhtvggsihhnrdgtohhmnecuvehluhhsthgvrhfuihiivgeptdenucfrrghrrghmpe hmrghilhhfrhhomhepghhrvggvnhhfohhosehuledvrdgvuh X-ME-Proxy: Received: by mail.messagingengine.com (Postfix) with ESMTPA; Tue, 30 Nov 2021 15:43:36 -0500 (EST) From: Fernando Ramos To: git@vger.kernel.org Cc: gitster@pobox.com, davvid@gmail.com, sunshine@sunshineco.com, seth@eseth.com, levraiphilippeblain@gmail.com, rogi@skylittlesystem.org Subject: [PATCH v3 0/3] vimdiff: new implementation with layout support Date: Tue, 30 Nov 2021 21:43:30 +0100 Message-Id: <20211130204333.174967-1-greenfoo@u92.eu> X-Mailer: git-send-email 2.34.0 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Precedence: bulk List-ID: X-Mailing-List: git@vger.kernel.org A few weeks ago I presented this RFC [1] where I introduced a new variant of the vimdiff merge tool ("vimdiff4") that creates three tabs (instead of just one) that look like this: ------------------------------------------ | | TAB #2 | TAB #3 | | ------------------------------------------ | | | | | LOCAL | BASE | REMOTE | | | | | <---- Same information ------------------------------------------ presented by the | | "standard" vimdiff | MERGED | merge tool | | ------------------------------------------ ------------------------------------------ | TAB #1 | | TAB #3 | | ------------------------------------------ | | | | | | | | | | BASE | LOCAL | <---- Only differences | | | between BASE and | | | LOCAL are shown | | | ------------------------------------------ ------------------------------------------ | TAB #1 | TAB #2 | | | ------------------------------------------ | | | | | | | | | | BASE | REMOTE | <---- Only differences | | | between BASE and | | | REMOTE are shown | | | ------------------------------------------ The motivation behind this was that, for non-trivial merges, the three way diff presented in the first tab tends to be very confusing and in these cases indivial diffs between BASE and LOCAL and between BASE and REMOTE are very useful. I have been using a "custom" merge tool for months to achieve this same result by adding these lines to my .gitconfig file: [mergetool "supermerge"] cmd = vim -f -d -c \"4wincmd w | wincmd J | tabnew | edit $LOCAL | vertical diffsplit $BASE | tabnew | edit $REMOTE | vertical diffsplit $BASE | 2tabprevious\" \"$LOCAL\" \"$BASE\" \"$REMOTE\" \"$MERGED\" trustExitCode = true ...and, because I found this "trick" very useful, I thought it would be a good idea to add it as a git built-in merge tool (called "vimdiff4" because 1, 2 and 3 had already been taken) for everyone to use... and that's exactly what the RFC I published did. Now... as you can see in the RFC thread [1], David and Juno suggested that maybe, instead of creating *yet another vimdiff variant*, we should take this opportunity to: * Come up with a more general way of defining arbitrary vim layouts. * Re-implement "vimdiff1", "vimdiff2" and "vimdiff3" using this new mechanism (after all, the only difference among them is that they present different layouts to the user) * Add documentation to all of this. And the result of that work is what I'm presenting today :) Some things I would like to mention: 1. There are three commits in this patch series: - The first one implements the logic to generate new arbitrary layouts and also re-defines "vimdiff1", "vimdiff2" and "vimdiff3" on top of it. - The second one adds documentation. It is probably a good idea to start reviewing this commit before the first one! - The last commit *is not meant to be merged now*. It removes "vimdiff1", "vimdiff2" and "vimdiff3", which is something that should only be done after one or two releases with a deprecation notice and only if everyone agrees to do so :) 2. "mergetools/vimdiff" is now a ~800 lines bash script, but most of it is documentation (which is embedded in the tool itself for easier maintenance) and unit tests. I have only tested it with bash, but I've tried not to use any command not already being used somewhere else, so I expect it to work in the same places it was working before (however, let me know if there are some shell compatibility requirements and I'll try to check them). 3. Regarding unit tests, "mergetool/vimdiff" contains instructions on how to run them (just call the script without arguments after making changes, to make sure you didn't break anything). Right now it prints "OK" on all test cases (obviously) [2] 3. The "git {diff,merge}tool --tool-help" command now also prints the documentation for each tool (instead of just its name, as before). You can see an example of the output here ([3] and [4]) Finally, let me say that, while I like what this patch series achieves, I would also *completely* understand if you decide not to merge it due to being a complex solution to a simple problem that can be solved (as I had been doing up until today) by just adding three line to one's .gitconfig. [mergetool "supermerge"] cmd = vim -f -d -c ...(custom complex sequence of vim commands)... trustExitCode = true Let me know what you think. Thanks. References: [1] https://lore.kernel.org/git/20211019212020.25385-1-greenfoo@u92.eu/#r [2] https://pastebin.com/kuQ5pETG [3] https://pastebin.com/yvLWxeiM [4] https://pastebin.com/qNc7qymp New in v2: * Remove deprecation of vimdiff{1,2,3} (they will remain for backwards compatibility, but built on top of the new generic layout mechanism) * Remove unnecessary "IFS=" * Replace DEBUG --> GIT_MERGETOOL_VIMDIFF_DEBUG * Stop using "local" (turns out it was not necessary) * Stop using bash arrays (use arrays of variables instead) * Stop using bash substring expansion (use a new "substring" function instead). * Refactor some of the internal loops to make the code faster. This was needed because the two previous changes (specially the one where I stop using "substring expansion") slowed down the script to a point where I had to wait for a few *seconds* before the "layout" string was resolved (recursion + bash forks are a recipe for sluggishness). Fortunately we are back to the low hundreds of milliseconds range. * Change markers: - File to save: * --> @ - New tab : ; --> + - Vert split : | --> , - Horz split : - --> / * Rewrite examples to use as few parenthesis as possible (typically zero) and better explain operators priority. * Other fixes to remove problems reported by "shellcheck --shell=sh" (which checks syntax agains the POSIX shell spec) * Rename "index_..." vars to make more obvious what they do. * Use "$" inside arithmetic expressions $((...)) NOTE: "shellcheck" issues a warning stating that "$" is not needed inside arithmetic expressions. * Use "test -n" instead of "! test -z" * Use "=" instead of "==" in "test" * Use "= 0" instead of "-eq 0" * Do not use "eval" nor "--" when copying a file to MERGED * Do not use "-o" with grep * Use <<-\EOF instead of <<\ENDOFMESSAGE and remove extra indent in "here docs". * Also, let me put here some answers to questions made in the replies to version v1 of this patch set: > What do backticks do in here? > > some_var=( > `# Name` "Rick Deckard" > `# Age` "Unknown" > `# Occupation` "Blade runner" > ) The backticks execute the code inside, which is a comment (everything after a "#" is considered a comment by the shell), thus it does nothing. The `# ...` trick can be used to insert inline comments in bash. Another example: $ ls -l `# long format` -h `# human readable` *.txt In any case, as you can see in this new revision, because I'm no longer using bash arrays, this trick to comment each test case is not needed anymore. > Why is "eval" needed here: > > eval "$merge_tool_path" \ > -f "$FINAL_CMD" "$LOCAL" "$BASE" "$REMOTE" "$MERGED" Variable "$FINAL_CMD" contains a string that looks like this: $ echo $FINAL_CMD -c "vim cmd 1 | vim cmd 2 | ..." -c "tabfirst" We need to call "vim" exactly like that, but if we do this... $ vim $FINAL_CMD ...then "vim" will be excev'ed with the following arguments: 1. -c 2. "vim 3. cmd 4. 1 ... ...instead of the desired scenario: 1. -c 2. cmd 1 | cmd 2 | ... 3. -c 4. tabfirst Using "eval" fixes this. A shorter example shows how this works: $ A="1 \"2 3\"" $ ls $A ls: cannot access '1': No such file or directory ls: cannot access '"2': No such file or directory ls: cannot access '3"': No such file or directory $ eval ls $A ls: cannot access '1': No such file or directory ls: cannot access '2 3': No such file or directory * Finally, I think I have addressed all comments to v1 *except for two things*: 1. Moving the documentation to "Documentation/" instead of having it embedded inside the "mergetools/vimdiff" script. 2. Move unit tests to the "test suit" For (1) I would like more details. Is the idea not to print any documentation when running "git mergetool --tool-help" and have all the details included in "git help mergetool" instead? Right now "git help mergetool" does not mention any tool specific details and, instead, redirects the user to "git mergetool --tool-help" (that's why I originally placed the new documentation there). In any case, all the doc has been placed on its own commit, so once we decide how to proceed its just a matter of reverting it. For (2) I still need to investigate how this is done and prepare a "v3" if needed :) New in v3: * Rebase on top of latest master * Documentation moved to "Documentation/". Running "git mergetool --tool-help" now works like before (ie. it just lists all available merge tools) except that a one line reference to a "man" entry is also printed for those entries providing one (right now, only vimdiff): $ git mergetool --tool-help 'git mergetool --tool=' may be set to one of the following: araxis meld vimdiff Run 'man git-mergetool--vimdiff' for details vimdiff1 Run 'man git-mergetool--vimdiff' for details ... * New test file "t/7609-mergetool--lib.sh" created. Right now it only contains one test which calls "mergetools/vimdiff" with a special environment set, causing layout unit tests to run. Fernando Ramos (3): vimdiff: new implementation with layout support vimdiff: add tool documentation vimdiff: integrate layout tests in the unit tests framework ('t' folder) Documentation/git-difftool--vimdiff.txt | 41 ++ Documentation/git-mergetool--vimdiff.txt | 195 ++++++++ git-mergetool--lib.sh | 14 + mergetools/vimdiff | 577 ++++++++++++++++++++++- t/t7609-mergetool--lib.sh | 17 + 5 files changed, 819 insertions(+), 25 deletions(-) create mode 100644 Documentation/git-difftool--vimdiff.txt create mode 100644 Documentation/git-mergetool--vimdiff.txt create mode 100755 t/t7609-mergetool--lib.sh base-commit: 35151cf0720460a897cde9b8039af364743240e7 -- 2.34.0