Disqus - Friends of yawar_amin

Re: Shortcomings of Mercurial

Jakub Narębski — Fri, 07 Mar 2008 05:53:32 -0000

A bit of Git advocacy:

* Git handles binary files correctly (it detect binariness using algorithm used by diff and other GNU tools, but you can mark given file as binary or text using gitattributes). Binariness of file does not matter to the Git libXdiff deltaification. Git can send / generate binary patches. Unfortunately, depending on binary file and how much it repesentation change due to small underlying change, Git can fail to detect file rename or file copying (it is based on similarity score).

* From some time Git ships with its own diff3 / rcsmerge file level merge tool. It would put CVS-like conflict markers in a file which has merge conflict. There is git-mergetool helper which helps to use configured graphical merge tool kile KDiff3, Meld, Emacs emerge etc.

* Git has core.sharedRepository configuration option, with possible values being: group, all. umask, to set correct permissions for git repository files, when using repository in shared fashion. Nevertheless DSCMs are usually geared to other workflows...

Re: Shortcomings of Mercurial

Jakub Narębski — Fri, 07 Mar 2008 15:16:49 -0000

@William Tanksley Jr:

First, when there is a merge conflict, Git leaves all three versions (base / ancestor, ours, theirs) in the index (commit staging area), and stores file with merge markers on filesystem. So you always have easy access ("git show :1:<file>" for example to get base version) to all versions.

Second, default pre-commit hook which ships with Git, and is by default installed in newly created repositories, detect merge markers in file (and also things like trailing whitespace) and refuses to make a commit if it looks like you didn't resolve a merge. You just have to enable it by making it executable ("chmod a+x .git/hooks/pre-commit"). Of course you can foce skipping verification phase if the tool misdetected merge markers with --no-verify option.

Third, you can use gitattributes to set merge driver for specified files, for example for all files, or/and you can use merge.default configuration option, choosing e.g. "binary" driver which leaves current version of file in the working dir in the case of conflicts.

BTW. why didn't you ask this question on Git mailing list or on #git channel on FreeNode? Git mailing list is open; you don't have to subscribe to post. Just make sure to specify that you'd like to get messages replied also to you, and/or read mailing list via mailing list archives or GMane NNTP (Usenet, news) interface.

Re: Shortcomings of Mercurial

Jakub Narębski — Fri, 07 Mar 2008 20:11:22 -0000

http://git.or.cz/gitwiki/Bl... see section "Portability" (Git on MacOS X).

As to why Git has so many dependencies: much of commands are written in either Perl (thus requiring Perl), or in shell scrips (thus requiring some of shell utils). Git also uses existing libraries, like libz, or libcurl, or libcrypto for things like compression, or HTTP protocol, or cryptographic hashes.

By the way, if I remember correctly, Mercurial has some of its code written in C for performance. And you have to got all those Python modules installed.

Re: Shortcomings of Mercurial

Jakub Narębski — Sat, 08 Mar 2008 14:17:19 -0000

@ William Tanksley Jr:

Errr... I think I haven't been clear enough. With Git you can either detect merge markers in (default, you only need to enable it) pre-commit hook and refusing commit unless overriden with --no-verify.

Or you can change default file-level merge driver for some files or for all files (using combination of gitattributes and config) from "text" (which inserts diff3 -E like merge markers) to "binary" (which leaves 'our' version on disk and does not insert any merge markers).

BTW: http://git.or.cz/gitwiki/Ma...

Re: You keep using that word “distributed”…

Jakub Narębski — Fri, 10 Oct 2008 11:21:11 -0000

I don't think hosting your own public repository over HTTP in Git is absurdly complicated (by the way, you can host via SSH, or via git protocol using git-daemon). You should share bare repository, and "dumb" HTTP protocol requires extra info present, which is provided by git-update-server-info, and which is automated by the use of update hook.

Re: When will Padre move to Git?

Jakub Narębski — Wed, 06 Jan 2010 19:30:08 -0000

But Git is what Perl 5 uses. And there is gitPAN.

Re: bitquabit - The fighting's been fun and all, but it's time to shut up and get along

Jakub Narębski — Thu, 11 Feb 2010 06:17:56 -0000

Is it the same as "the Hg-Git mercurial plugin" http://hg-git.github.com/

Re: bitquabit - The fighting's been fun and all, but it's time to shut up and get along

Jakub Narębski — Thu, 11 Feb 2010 06:25:05 -0000

Git from version 1.7.0 would (by default) refuse to push into currently checked out branch.
Also all tutorials, introductions, manuals, etc. tell you to not push into non-bare repository.

By default means that you can configure this behavior.

Re: bitquabit - The fighting's been fun and all, but it's time to shut up and get along

Jakub Narębski — Thu, 11 Feb 2010 06:34:15 -0000

A bit of self promotion: with the help from Steve Losh (and his blog entry[1]) and people on #mercurial IRC channel on FreeNode for the Mercurial side of story, myself writing the Git part o story, there is nice writeup on differences between Git and Mercurial on Stackoverflow:

"Git and Mercurial - Compare and Contrast"
http://stackoverflow.com/qu...

[1] http://stevelosh.com/blog/e...

Re: bitquabit - Kiln's Evolution, Part 2: From Prototype to Beta

Jakub Narębski — Thu, 11 Feb 2010 10:51:40 -0000

How Kiln compares to Gerrit Code Review: http://code.google.com/p/ge... (demo at http://review.source.androi...)?

Re: The Real Difference Between Mercurial and Git / Steve Losh

Jakub Narębski — Sat, 13 Feb 2010 07:41:31 -0000

About usefulness of git index: even if you don't use commit step by step (for 'git diff' to show only changes you are working on), and don't create partial commits / commits with dirty tree (using 'git stash --keep-index' to test them), git index is dead useful when merging... especially if you bork *part* of merge.

See "In praise of git’s index" by Aristotle Pagaltzis: http://plasmasturm.org/log/...
(found via http://tomayko.com/)

Re: The Real Difference Between Mercurial and Git / Steve Losh

Jakub Narębski — Sat, 13 Feb 2010 12:32:33 -0000

Patch management interface, such as Mercurial mq, or StGIT (Stacked GIT) or Guilt for Git, can help only with the first part, to commit step by step (by refreshing the patch after each step).

It wouldn't help with creating commit from a dirty tree (to split changes into many commits), at least not directly. And it wouldn't be of help when you did mistake resolving some conflict (but some other conflicts were resolved correctly, and some other conflicts are yet to be resolved).

Re: bitquabit - Kiln's Evolution, Part 2: From Prototype to Beta

Jakub Narębski — Tue, 23 Feb 2010 06:48:59 -0000

FYI, here is link to Shawn O. Pearce (spearce), one of Gerrit developers, comparing Gerrit and Kiln on #git IRC channel:

http://colabti.de/irclogger...

for one, [Gerrit] is not commercial. :-)

kiln is hg specific, gerrit is git specific. kiln seems to work under the notion of everyone having their own fork repositories, gerrit works under then notion of a shared team repository with all reviews pending in that.

but, hey, competition is good. glad to see something come along for hg.

i'm pretty sure fogcreek and myself approached the problem from different angles. i approached from the android side where we have >200+ git repositories and users don't want to fork all of them just to upload a 1 line code change. so i went with everything shared in one project. they seem to have approached from the more typical DVCS case of 1 repository : 1 project, with developers having only a few projects.

Re: bitquabit - Kiln's Evolution, Part 2: From Prototype to Beta

Jakub Narębski — Tue, 23 Feb 2010 17:08:53 -0000

Discussion continues (spearce, tghw, bmp) on #git:

http://colabti.org/irclogge...

Re: Running Processes

Jakub Narębski — Mon, 01 Mar 2010 09:43:03 -0000

What about init-ng?

Re: Sam Graham's website > Perlblog > Template::Benchmark.pm v1.00 Released

Jakub Narębski — Mon, 10 May 2010 15:03:19 -0000

Can Template::Benchmark measure latency (i.e. perceived snapiness of templating engine), and not only throughput (end-to-end performance)? Is automatical exaping of HTML one of features of templating engines that is considered?

Re: Sam Graham's website > Perlblog > Template::Benchmark.pm v1.00 Released

Jakub Narębski — Mon, 10 May 2010 19:26:52 -0000

By difference between latency and throughput I mean here something like difference between 'time run_tempate | head >/dev/null' and 'time run_template >/dev/null'. Here it would be measuring time it takes for templating engine to get first few lines of output, or get first page of output.

Templating engine might have good measured performance, but do not feel snappy.

As for (semi)automatic HTML escaping, I think it is one of features of templating engines. Some are geared towards generating HTML and have escaping built in (Mojo::Template's '%==', automatic HTML escaping in HTML::Zoom), some have HTML escaping filters (e.g. Mason and Template Toolkit).

When one wants to select templating engine to be used in a web application, easy HTML escaping should be one of features taken into account, I think.

Re: Sam Graham's website > Perlblog > Template::Benchmark.pm v1.00 Released

Jakub Narębski — Tue, 11 May 2010 06:17:44 -0000

Do you plan on adding plugins for other Perl templating systems, like Perlish templates Template::Declare and similar Markapl, selector based templates with pure (X)HTML/XML as template Template::Semantic and HTML::Zoom (and pQuery), and perhaps also TAL templates (from Zope CMS) Template::TAL and Petal, Haml templates (from Ruby) Text::Haml?

Re: Sam Graham's website > Perlblog > Perl Template Roundup Announced

Jakub Narębski — Thu, 01 Jul 2010 05:14:49 -0000

In "Template Engines Used for the June 2010 Roundup" both TeMMHM and TeMMTeTe have the same description: 'Text::MicroMason (2.07) using Text::MicroMason::HTMLMason (no version number)'

Re: Sam Graham's website > Perlblog > Perl Template Roundup Announced

Jakub Narębski — Thu, 01 Jul 2010 12:58:42 -0000

A few possible improvements: "Reports" page, in "Choose a report" table has first column with uncached_string..instance_reuse unnamed (without column header). It is Cache Types.

The "Jump to chart:" menu in individual per-cache-type reports, and also headers for individual per-feature results, could use short description of each feature, e.g. <tt>literal_text - A chunk of literal text, dumped through to the output largely unchanged from its form in the template.</tt> The plots would be more readable if they were horizontal rather than vertical, but then I guess there would be problem with describing bins (describing axis). The x-axis (the value axis) is not described; is in seconds, or is it is operations per seconds? There really should be mentioned that lower / higher is better (as it suits).

The summary page also doesn't have description if lower or higher means better performance.

Re: Sam Graham's website > Perlblog > Perl Template Roundup Announced

Jakub Narębski — Mon, 05 Jul 2010 08:56:24 -0000

One more thing, although I am not sure if you wouldn't hit Google Charts limitation: in every place where you use shortcut name for caching engine + behaviour, e.g. HTP you could put full name, e.g. HTML::Template::Pro (0.95), in 'title' attribute of element (or <abbr> element) to be visible on mouseover. This way one wouldn't need to go back to starting page if one forgets e.g. what HTP is shortcut for.

Re: Sam Graham's website > Perlblog > Perl Template Roundup Announced

Jakub Narębski — Mon, 05 Jul 2010 10:35:00 -0000

Sam Graham:

Hmm, there should be a tooltip mouse-over for the charts giving that information already. Is that not working for you?

I am sorry, I haven't checked if it was corrected before writing last comment. I see tooltip with all the shortcuts for chart as a whole; I would prefer having small tooltip over only single shortcut, but that might be not possible with Google Chart API.

Re: Hacking Thy Fearful Symmetry

Jakub Narębski — Sun, 14 Nov 2010 07:23:53 -0000

What about adding this functionality to cpan-mangler?

Re: Hacking Thy Fearful Symmetry

Jakub Narębski — Sun, 06 Mar 2011 06:24:13 -0000

Does it / would it allow one to recant or change one's vote?

A good example of implemented voting mechanizm can be Reddit, or Stackoverflow.

Re: My Git Workflow and Why You Should Use Git

Jakub Narębski — Wed, 14 Dec 2011 11:21:22 -0000

What do those do?

$ git cl upload
$ git try

These are not git commmands