I'm working on a new weave-based data structure for storing version control history. This will undoubtedly cause some religious wars about whether it's The Right Way Of Doing Things when it comes out, but that isn't my question right now.
My question has to do with what output blame should give. When a line of code has been added, removed, and merged into itself a number of times, it isn't always clear what revision should get blame for it. Notably this means that when a section of code is deleted, all records of it having been there is gone, and there is no blame for the removal. Everyone I've gone over this issue with has said that trying to do better simply isn't worth it. Sometimes people put in the hack that the line after the section which got deleted has its blame changed from whatever it actually was to the revision when the section got deleted. Presumably if the section is at the end then the last line get its blame changed, and if the file winds up empty then the blame really does disappear into the aether, because there's literally nowhere left to put blame information. For various technical reasons I won't be using this hack, but assume that continuing but with this completely undocumented but de facto standard practice will be uncontroversial (but feel free to flame me and get it out of your system).
Moving on to my actual question. Usually in blame for each line you look at the complete history of where it was added and removed in the history and using three-way merge (or, in the case of criss-cross merges, random bullshit) and based on the relationships between those you determine whether the line should have been there based on its history, and if it shouldn't but is then you mark it as new with the current revision. In the case where a line occurs in multiple ancestors with different blames then it picks which one to inherit arbitrarily. Again, I assume that continuing with this completely undocumented but de facto standard practice will be uncontroversial.
Where my new system diverges is that rather than doing a complicated calculation of whether a given line should be in the current revision based on a complex calculation of the whole history, it simply looks at the immediate ancestors, and if the line is in any of them it picks an arbitrary one to inherit the blame from. I'm making this change for largely technical reasons (and it's entirely possible that other blame implementations do the same thing, for similar technical reasons and a lack of caring) but after thinking about it a bit part of me actually prefers the new behavior as being more intuitive and predictable than the old one. What does everybody think?
I actually wrote a one of the blame implementations out there (Subversion's current one I believe, unless someone replaced it in the past year or two). I helped with some others as well.
At least most implementations of blame don't do what you describe:
Usually in blame for each line you look at the complete history of where it was added and removed in the history and using three way merge (or, in the case of criss-cross merges, random bullshit) and based on the relationships between those you determine whether the line should have been there based on its history, and if it shouldn't but is then you mark it as new with the current revision. In the case where a line occurs in multiple ancestors with different blames then it picks which one to inherit arbitrarily. Again, I assume that continuing with this completely undocumented but de facto standard practice will be uncontroversial.
Actually, most blames are significantly less complex than this and don't bother trying to use the relationships at all, but they just walk parents in some arbitrary order, using simple delta structures (usually the same internal structure whatever diff algorithm they have uses before it turns it into textual output) to see if the chunk changed, and if so, blame it, and mark that line as done.
For example, Mercurial just does an iterative depth first search until all lines are blamed. It doesn't try to take into account whether the relationships make it unlikely it blamed the right one.
Git does do something a bit more complicated, but still, not quite like you describe.
Subversion does what Mercurial does, but the history graph is very simple, so it's even easier.
In turn, what you are suggesting is, in fact, what all of them really do:
Pick an arbitrary ancestor and follow that path down the rabbit hole until it's done, and if it doesn't cause you to have blamed all the lines, arbitrarily pick the next ancestor, continue until all blame is assigned.
On a personal level, I prefer your simplified option.
Reason: Blame isn't used very much anyway.
So I don't see a point in wasting a lot of time doing a comprehensive implementation of it.
It's true. Blame has largely turned out to be one of those "pot of gold at the end of the rainbow" features. It looked really cool from those of us standing on the ground, dreaming about a day when we could just click on a file and see who wrote which lines of code. But now that it's widely implemented, most of us have come to realize that it actually isn't very helpful. Check the activity on the blame tag here on Stack Overflow. It is underwhemingly desolate.
I have run across dozens of "blame-worthy" scenarios in recent months alone, and in most cases I have attempted to use blame first, and found it either cumbersome or utterly unhelpful. Instead, I found the information I needed by doing a simple filtered changelog on the file in question. In some cases, I could have found the information using Blame as well, had I been persistent, but it would have taken much longer.
The main problem is code formatting changes. The first-tier blame for almost everything was listed as... me! Why? Because I'm the one responsible for fixing newlines and tabs, re-sorting function order, splitting functions into separate utility modules, fixing comment typos, and improving or simplifying code flow. And if it wasn't me, someone else had done a whitespace or block-move somewhere along-the-way as well. In order to get a meaningful blame on anything dating back to a time before I can already remember without the help of blame, I had to roll back revisions and re-blame. And re-blame again. And again.
So in order for a blame to actually be a useful time saver for more than the most lucky of situations, the blame has to be able to heuristicly make its way past newline, whitespace, and ideally block copy/move changes. That sounds like a very tall order, especially when scouring the changelog for a single file, most of the time, it won't yield many diffs anyway and you can just sift through by hand fairly quickly. (The notable exception being, perhaps, badly engineered source trees where 90% of the code is stuffed in one or two ginormous files... but who these days in a collaborative coding environment does much of that anymore?).
Conclusion: Give it a bare-bones implementation of blame, because some people like to see "it can blame!" on the features list. And then move on to things that matter. Enjoy!
The line-merge algorithm is stupider than the developer. If they disagree, that just indicates that the merger is wrong rather than indicating a decision point. So, the simplified logic should actually be more correct.
Related
For text, not source code, files like LaTeX, markdown, restructuredText, usually single line breaks does not matter for the semantics and they are frequently refilled within 80 columns. When things are changed, the line break might change by quite a lot. So the common line-by-line diff and patch tools do not actually work for them very well. So I am wondering if there already exist good tools for diffing, patching and even merging this kind of changes? wdiff and git diff --color-words does exactly the kind of thing, but they seem to lack the patching and merging capability. Ideally, if we have got a line
He do not owe us nothing.
and one author changed it into
He do not owe us anything.
and another author changed it into
He does not owe us nothing.
then a merge could give
He does not owe us anything.
without conflict. That is the ideal result. Thanks in advance.
Besides meld, you can also use Beyond Compare or WinMerge
I have a lot of HTML files (10,000's and GBs worth) scraped from a server and I want to check to make sure the server produces the same results after some modifications but ignore kinds of differences that don't matter, e.g. whitespace, missing newlines, timestamps, small changes in some kinds of number, etc.
Does anyone know of a tool for doing this? I'd really rather not do more filtering than I have to.
(Oh and it needs to run under linux)
You might consider using a clone detector such as our CloneDR. This tool parses large sets of computer program (HTML is special case) files, builds abstract syntax trees representing the essential structure of each files, and compares programs for similarity.
Because it is comparing essential program structure, it ignores inessential differences such as comments and whitespace, and deterimines that two code segments are either identical or one can be obtained from the other by substituting other blocks of code. The latter allows the recognition of code that has been modified in various ways. You can see samples of clone detection runs on a variety of computer languages at the web site.
In your case, what you would be looking for are files in system A which are essentially clones (exact or near misses) of files in system B. As a general rule, if a file a is a variant of file b (e.g., with a few changes) the CloneDr will report it as a clone and show the exact differences.
At the scale of 20,000 files, I can see why you want a tool, and I can see why you want near-miss matches rather than exact matches.
Doesn't run under Linux, but I assume your problem is hard to enough to solve so that isn't what you are optimizing.
I use winmerge alot in windows and from what i can see some people enjoy meld in linux, so perhaps that could do the trick for you
http://meld.sourceforge.net/
Other examples i saw from a quick googling was Kompare,xxdiff.sourceforge.net, and kdiff3.sourceforge.net
(could only post 1 link so wrote the adresses to xxdiff and kdiff3 as text)
Beyond Compare is purchased software that is actually worth the money (I never thought I'd hear myself typing that!). It is GUI based but handles thousands of files very well. It will allow you to specify unimportant changes with regular expressions as well as whitespace (beginning, middle and end of line). The feature set is very extensive, check out a trial download.
I do not work for this company, I just use Beyond Compare every day at work and enjoy it every time!
What source control products have a "diff" facility that ignores white space, braces, etc., in calculating the difference between checked-in versions? I seem to remember that Clearcase's diff did this but Visual SourceSafe (or at least the version I used) did not.
The reason I ask is probably pretty typical. Four perfectly reasonable developers on a team have four entirely different ways of formatting their code. Upon checking out the code last changed by someone else, each will immediately run some kind of program or editor macro to format things the way they like. They make actual code changes. They check-in their changes. They go on vacation. Two days later that program, which had been running fine for two years, blows up. The developer assigned to the bug does a diff between versions and finds 204 differences, only 3 of which are of any significance, because the diff algorithm is lame.
Yes, you can have coding standards. Most everyone finds them dreadful. A solution where everyone can have their cake and eat it too seems far more preferable.
=========
EDIT: Thanks to everyone for some great suggestions.
What I take away from this is:
(1) A source control system with plug-in type diffs is preferable.
(2) Find a diff with suitable options.
(3) Use a good source formatting program and settle on a check-in standard.
Sounds like a plan. Thanks again.
Git does have these options:
--ignore-space-at-eol
Ignore changes in whitespace at EOL.
-b, --ignore-space-change
Ignore changes in amount of whitespace. This ignores whitespace at line end, and considers all other sequences of one or more
whitespace characters to be equivalent.
-w, --ignore-all-space
Ignore whitespace when comparing lines. This ignores differences even if one line has whitespace where the other line has
none.
I am not sure if brace changes can be ignored using Git's diff.
If it is C/C++ code, you can define Astyle rules and then convert the source code's brace style to the one that you want, using Astyle. A git diff will then produce sane output.
Choose one (dreadful) coding standard, write it down in some official coding standards document, and get on with your life, messing with whitespace is not productive work.
And remember you are a professional developer, it's your job to get the project done, changing anything in the code because of a personal style preference hurts the project - it wont only make diff-ing more difficult, it can also introduce hard to find problems if your source formatter or compiler has bugs (and your fancy diff tool won't save you when two co-worker start fighting over casing).
And if someone just doesn't agree to work with the selected style just remind him (or her) that he is programming as a profession not as an hobby, see http://www.ericsink.com/entries/No_Great_Hackers.html
Maybe you should choose one format and run some indentation tool before checking in so that each person can check out, reformat to his/her own preferences, do the changes, reformat back to the official standard and then check in?
A couple of extra steps but they already use indentation tools when working. Maybe it can be a triggered check-in script?
Edit: this would perhaps also solve the brace problem.
(I haven't tried this solution myself, hence the "perhapes" and "maybes", but I have been in projects with the same problems, and it is a pain to try to go through diffs with hundreds of irrelevant changes that are not limited to whitespace, but includes the formatting itself.)
As explained in Is it possible for git-merge to ignore line-ending differences?, it is more a matter to associate the right diff tool to your favorite VCS, rather than to rely on the right VCS option (even if Git does have some options regarding whitespace, like the one mentioned in Alan's answer, it will always be not as complete as one would like).
DiffMerge is the more complete on those "ignore" options, as it can not only ignore spaces but also other "variations" based on the programming language used in a given file.
Subversion apparently supports this, either natively in the latest versions, or by using an alternate diff like Gnu Diff.
Beyond Compare does this (and much much more) and you can integrate it either in Subversion or Sourcesafe as an external diff tool.
At least two brilliant programmers, Linus Torvalds and Guido von Rossum, disparage the practice of putting keywords into a file that expand to show the version number, last author, etc.
I know how keyword differences clutter up diffs. One of the reasons I like SlickEdit's DiffZilla is because it can be set to skip leading comments.
However, I have vivid memories of team-programming where we had four versions of a file (two different releases, a customer one-off, and the development version) all open for patching at the same time, and was quite helpful to verify with a glance that each time we navigated to an included header we got the proper one, and each time we pasted code the source and destination were what we expected.
There is also the where-did-this-file-come-from problem that arises when a hasty developer copies a file from one place to another using the file system, rather than checking it out of the repository using the tool; or, more defensibly, when files under control in locations A, B, and C need to be marshalled (with cherry-picking) into a distribution location D.
In places where VCS keywords are banned, how do you cope?
I've never used VCS keywords in my entire career, over 30 years. From the most primitive VCS system I've used, up to the present (TFS), I've used some other structure to understand "where I am".
I am rarely in a situation where I've only got one file to work with. I've usually got all the other files necessary to build the project or set of projects. I usually use branching (or streams on one occasion), and I'm working on some slice of the given branch or stream.
If I'm working on multiple branches or streams, I'll have one directory tree for each. All I need to do to know what file I'm working on is check the file path, at the very worst.
At the very best, the version control system will tell you exactly which version of the file you're working on, what the change history is, who else is working on different versions of the file, and anything else you'd care to know.
This doesn't exactly answer your question, but I imagine Linus and Guido have reasons for disliking keywords that don't apply to small-team corporate development.
An $Id$ tag for instance, has what you could consider to be a global version number. Linux and I guess also Python development is fragmented enough that no number can be global. Lots of people have their own repositories all over the place that would fill in their own $Id$ values and then those patches might be sent to Linus or Guido's repositories where they don't make any sense.
However, in your environment, you probably have one central repository which would assign these and it would be fine. Sounds like you're using git. I wonder if it's possible to configure the central git repository to do tag substitution while the local developer repositories don't. Or perhaps it's better to get the commit hash in the tag.
I seem to get into an annual debate about the use of the $Log$ keyword. My point of view is this:
$Log$ is white hot death.
All it does is jam marginally relevant spam into your source files. Any information that anyone thinks they might be able to get from a $Log$ is more readily available from (and is likely to be more accurate in) your version control system.
So, here's the question: how would you explain to an "old school" coder (who thinks that $Log$ is the way to manage source code changes) that we have better tools now?
The CVSNT remarks on $Log$ are a good start but they're just not pointed enough. To date, the closest that I've come to a one-liner that I've managed to come up with is "$Log$ is a wish. You're hoping that what gets spammed into your file has any relation to what really happened to this file."
PS for clarity: when I say "old school," I mean old in attitude, not old in years. My first programming paycheck (and a remarkably modest one it was, too) was sometime in 1986 and I never thought $Log$ was a good idea.
I think the Subversion FAQ also has a good explanation.
$Log$ is a total horror the moment you start merging changes
between branches. You're practically guaranteed to get conflicts there,
which -- because of the nature of this keyword -- simply cannot be
resolved automatically.
In addition to what the others have said, try putting a comment (/* ... */) into a commit message :->.
The amount of useful bits in a source file slowly decreases as changes are made to it with that $Log$ statement in it. We had it in some files that came from CVS and number of lines of $Log$ statements was on the order of 10x longer than the executable code in file actually was. And it had a few groups of duplicates caused by bad merging from some branches.
You may consider (emphasis on may) embedding immutable meta-data in your file.
(See the debate between me and an "older schooler" : Embedded Version Numbers - Good or Evil?).
Even though I have always considered that practice as evil (mixing meta-data information into data), introducing "merge hell", one could argue that it could work, with the right merge manager, for immutable meta-data with a fixed format, like:
$Revision$ $Revision: 9.13 $
$Date$ $Date: 2009/03/06 06:52:26 $
$RCSfile$ $RCSfile: stderr.c,v $
But mutable meta-data like logs ? With unknown format or content ? That is bound to fail.