file page 36 - Dave's Blog

Which which - Batch File Hackiness

2007 Aug 9, 5:41To satisfy my hands which have already learned to type *nix commands I like to install Win32 versions of common GNU utilities. Unfortunately, the which command is a rather literal port and requires you to enter the entire name of the command for which you're looking. That is 'which which' won't find itself but 'which which.exe' will. This makes this almost useless for me so I thought to write my own as a batch file. I had learned about a few goodies available in cmd.exe that I thought would make this an easy task. It turned out to be more difficult than I thought.

for /F "usebackq tokens=*" %%a in ( `"echo %PATH:;=& echo %"` ) do (
    for /F "usebackq tokens=*" %%b in ( `"echo %PATHEXT:;=& echo %"` ) do (
        if exist "%%a"\%1%%b (
            for  %%c in ( "%%a"\%1%%b ) do (
                echo %%~fc
            )
        )
    )
)

The environment variables PATH and PATHEXT hold the list of paths to search through to find commands, and the extensions of files that should be run as commands respectively. The 'for /F "usebackq tokens=*" %%a in (...) do (...)' runs the 'do' portion with %%a sequentially taking on the value of every line in the 'in' portion. That's nice, but PATH and PATHEXT don't have their elements on different lines and I don't know of a way to escape a newline character to appear in a batch file. In order to get the PATH and PATHEXT's elements onto different lines I used the %ENV:a=b% syntax which replaces occurrences of a with b in the value of ENV. I replaced the ';' delimiter with the text '& echo ' which means %PATHEXT:;=& echo% evaluates to something like "echo .COM& echo .EXE& echo .BAT& ...". I have to put the whole expression in double quotes in order to escape the '&' for appearing in the batch file. The usebackq and the backwards quotes means that the backquoted string should be replaced with the output of the execution of its content. So in that fashion I'm able to get each element of the env. variable onto new lines. The rest is pretty straight forward.

Also, it supports wildcards:

C:\Users\davris>which.cmd *hi*
C:\Windows\System32\GRAPHICS.COM
C:\Windows\System32\SearchIndexer.exe
D:\bin\which.exe
D:\bin\which.cmd

which cmd technical batch for

Two-for Script File

2007 Aug 6, 5:40I was messing with the XSLT to XSL Converter source which is a javascript file that can be run with cscript.exe. I've changed it to be like a very basic version of xsltproc that simply runs an XML file through an XSLT. I also wanted to run this from the command prompt without writing "cscript ..." everytime. I decided to make like perl programmers I've seen and make a JS file that works as a batch file and a JS file at the same time.

Here's a basic version of what I ended doing applied to a 'hello world' script named helloworld.cmd:

/* 2> NUL
@echo off
cscript /e:javascript /nologo "%~f0" %*
@goto :eof

    Hello World
        Says 'Hello world.' when you run it.
*/

var outText = 'Hello world.';
WScript.Echo(outText);

Running this on a command prompt gives the following:

C:\Users\davris>helloworld

C:\Users\davris>/*  2>NUL
Hello world.

However, after a little more experimentation I found this was slightly overkill for my purposes since if I rename the file to helloworld.js and just type its name like a command it is run by cscript:

C:\Users\davris>helloworld
Microsoft (R) Windows Script Host Version 5.7
Copyright (C) Microsoft Corporation. All rights reserved.

Hello world.

So this time I didn't need all that but if ever in the future I need to run a batch file then a JS file I can do it with one file...cmd js technical cscript batch xslt xsl javascript

Wp64 Issues

2007 Aug 6, 3:43Miladin told me about the Visual Studio compiler's promising option Wp64 that finds 64bit portability issues when compiling in 32bit. If, for instance, you cast from a (long*) to a (long) you get a W4 warning. However, the #defines are still set for 32bit builds. This means that other parts of the code can make assumptions based on the #defines that are valid on 32bit but generate 64bit errors or warnings.

For instance, in winuser.h the public published Windows header file there's the following:

...
#ifdef _WIN64
...
WINUSERAPI
LONG_PTR
WINAPI
SetWindowLongPtrA(
    __in HWND hWnd,
    __in int nIndex,
    __in LONG_PTR dwNewLong);
...
#else  /* _WIN64 */
...
#define SetWindowLongPtrA   SetWindowLongA
...
#endif /* _WIN64 */
...

In 64bit everything's normal but in 32bit SetWindowLongPtrA is #defined to SetWindowLongA which takes a LONG rather than a LONG_PTR. So take the following code snippet:

...
LONG_PTR inputValue = 0;
LONG_PTR error = SetWindowLongPtrA(hWnd, nIndex, inputValue);
...

This looks fine but generates warnings with the Wp64 flag.

In 64 bit, p is cast to (LONG_PTR) and that's great because we're actually calling SetWindowLongPtrA which takes a LONG_PTR. In 32 bit, p is cast to (LONG_PTR) which is then implicitly cast to (LONG) because we're actually calling SetWindowLongA. LONG and LONG_PTR are the same size in 32bit which is fine but if you turn on the Wp64 flag there's a W4 warning because of the implicit cast from a larger size to a smaller size if you were to compile for 64bit. So even though doing a 32bit or 64bit compile would have worked just fine, if you turn on the Wp64 flag for 32bit you'd get an error here.

It looks like I'm the most recent in a list of people to notice this issue. Well I investigated this so... I'm blogging about it too!wp64 technical 64bit compiler c++ visual-studio setwindowlongptra

Backup Notes

2007 Jul 13, 8:30I bought an external backup drive a few weekends ago. I've previously setup a Subversion repository so I decided to move everything into the repository and then back it up. So in went the contents of all of my %USERPROFILE% and ~ directories with a bit of sorting and pruning. Not too much though given its much easier to dump in everything and search for what I want then to take the time to examine and grade each file. What follows are the notes I took while setting this up. It takes me a bit of time to look up the help on each command so I figure I'll write it all down here for the benefit of myself and potentially others...

Setting Up the Backup Drive For Linux
I first changed the filesystem on the drive to ext3. I plugged it into my USB2.0 port and ran fdisk:

sudo fdisk /dev/sda

Useful commands I used to do this follow mostly in order:

m

help

p

print current partitions

d

delete current partition

n

create new partition (I used the defaults)

w

write changes and exit

Then I formatted for ext3.

sudo mkfs.ext3 /dev/sda1

I made it easy to mount:

sudo vim /etc/fstab
# added line to end:
/dev/sda1 /media/backup ext3 rw,user,noauto 0 0

I setup the directory structure on the disk

mount /media/backup
sudo mkdir /media/backup/users
sudo mkdir /media/backup/users/dave
sudo chown dave:dave /media/backup/users/dave

After all that its easy to make a copy of the Subversion repository:

mount /media/backup
cp -Rv /home/dave/svn /media/backup/users/dave/
umount /media/backup

Next on the agenda is to add a cron job to do this regularly.

Subversion Command Reference
On a machine that has local access to the Subversion repository you can check out a specific subdirectory as follows using the file scheme:

svn co file:///home/dave/svn/trunk/web/dave%40deletethis.net/public_html

Note also that although one of my directories is named 'dave@deletethis.net' Subversion requires the '@' to be percent-encoded.
Other useful subversion commands:

svn help

help

svn list file:///home/dave/svn/

list all files in root dir of svn depot

svn list -R file:///home/dave/svn/

list all files in svn depot

svn list -R file:///home/dave/svn/ | grep \/$

list all directories

svn status

List status of all files in the working copy directory as in - modified, not in repository, etc

svn update

Brings the working copy up to date wrt the repository

svn commit

Commit changes from the working copy to the repository

svn add / move / delete

Perform the specified action -- occurs immediately

Setting up Windows Client for Auto Auth into SVN
When using an SVN client on Windows via svn+ssh its useful to have the Windows automatically generate connections to the SVN server. I use putty on my Windows machines so I read the directions on using public keys with putty.

putty.exe dave@deletethis.net
cd .ssh
vim authorized_keys # leave the putty window open for now
puttygen.exe
Click the 'generate' button
Move the mouse around until finished
Copy text in 'Public key for pasting into OpenSSH authorized_keys file:' to putty window & save & close putty window
Enter Key passphrase & Comment in puttygen
Save the private key somewhere private
pageant.exe
'Add Key' the private key just saved.

Checking out using Tortoise SVN
On one of my Windows machines I've already installed Tortoise SVN. Checking out from my SVN repository was really easy. I just right clicked in Explorer in a directory and selected "SVN Checkout...". Then in the following dialog I entered the svn URI:

svn+ssh://dave@deletethis.net/home/dave/svn/trunk/web/dave%40deletethis.net/public_html/

Note again that the '@' that is part of the directory name is percent-encoded as '%40' while the '@' in the userinfo is not.

Windows Command Line Check Out
On my media center I didn't want to install Tortoise SVN so rather I used the command line tool. I setup pageant like before the only difficulty was getting the SVN command line tool to use putty. With the default configuration you can use the SVN_SSH environment variable to point at a compliant SSH command line tool. The trick is that its interpreted as a backslash escaped string. So I set mine thusly:

set SVN_SSH=C:\\users\\dave\\bin\\putty\\plink.exe

The escaping solved the vague error I received about not being able to create the tunnel.backup technical personal windows svn linux subversion

Ozzie

2007 Jun 25, 3:13I keep seeing 'Ozzie' on emails and such now due mainly to Ray Ozzie who is now the Chief Software Architect at Microsoft and his brother Jack Ozzie. Whenever I see his name I think of Ozzie from Chrono Trigger. He was one third of a trio of villains, the other two being Flea and Slash. I feel like I should be thinking of the Ozzy for which this Ozzie was named but I really don't.

My next thought on Ozzie is the Scottish guy who went to my high school. He'd shout 'Ozzie! Ozzie! Ozzie!' to which listeners were compelled to respond 'Oi! Oi! Oi!'. The wikipedia article on the chant has some thoughts on the origins but I suppose at Microsoft it could take on entirely new meaning. I really hope I'm someday in a meeting with Ray or Jack Ozzie and have the opportunity...ozzy personal ozzie random nontechnical

Office Remodel

2007 Jun 11, 4:20My manager has come back this week from a 10 week vacation and paternity leave. In response and similar to other office hacks some dedicated coworkers and I decided to do something to my manager's office.

While gone we knew my manager, Venkat, was getting into meditation. My coworker Vishu had the excellent idea of easing Venkat back into work by making his office better suited for meditation. To start with, we updated his nameplate with an Om.

Next we emptied his office of anything that could distract him from meditation and replaced it with a yoga mat. Of course I left a copy of the specification for the remodel in his office.microsoft personal office humor nontechnical

IE7 Feed Display Update

2007 May 22, 3:22I've created an update to the IE7 feed display.

After working on my update to the XML source view I tried running my resourcelist program on other IE DLLs including ieframe. I found that one of the resources in ieframe is the XSLT used to turn an RSS feed into the IE7 feed display.

My first thought for this was that I could embed enclosures into the feed display. For instance, have controls for youtube.com videos or podcast audio files directly in the feed display. However, I found that I can't use object or embed tags that rely on ActiveX controls in the page or in frames in the feed display.

With that through I decided I could at least add support for some RSS extensions. Thanks to IE7's RSS platform which provides a normalized view of RSS feeds it was really easy to do this. I went to several popular RSS feeds and RSS feeds that I like and took a look at the source to see what extensions I might want to add support for.

For digg.com I added support for their RSS extension which includes digg count, and submitter name and icon. I added the digg count in a box on the right and tried to make it fit in stylistically. For the iTunes RSS extension I add the feed icon, feed author, and descriptions. I was surprised by how much of the podcasts content was missing from the feed view. I also added support for a few other misc things: the slash RSS extension's section and department, the feed description to the top of the feed display, and the atom author icon.

I wonder what other goodies lurk in IE's resources...feed res slashdot digg resource itunes technical browser ie rss extension

New XSLT - IE7 XML Source View Upgrade Part 2

2007 May 11, 8:55Last time, I had written some resource tools to allow me to view and modify Windows module resources in my ultimate and noble quest to implement the XML content-type fragment in IE7. Using the resource tools I found that MSXML3.DLL isn't signed and that I can replace the XSLT embedded resource with my own, which is great news and means I could continue in my endevour. In the following I discuss how I came up with this replacement for IE7's XML source view.

At first I thought I could just modify the existing XSLT but it turns out that it isn't exactly an XSLT, rather its an IE5 XSL. I tried using the XSL to XSLT converter linked to on MSDN, however the resulting document still requires manual modification. But I didn't want to muck about in their weird language and I figured I could write my own XSLT faster than I could figure out how theirs worked.

I began work on the new XSLT and found it relatively easy to produce. First I got indenting working with all the XML nodes represented appropriately and different CSS classes attached to them to make it easy to do syntax highlighting. Next I added in some javascript to allow for closing and opening of elements. At this point my XSLT had the same features as the original XSL.

Next was the XML mimetype fragment which uses XPointer, a framework around various different schemes for naming parts of an XML document. I focused on the XPointer scheme which is an extended version of XPath. So I named my first task as getting XPaths working. Thankfully javascript running in the HTML document produced by running my XSLT on an XML document has access to the original XML document object via the document.XMLDocument property. From this this I can execute XPaths, however there's no builtin way to map from the XML nodes selected by the XPath to the HTML elements that I produced to represent them. So I created a recursive javascript function and XSLT named-template that both produce the same unique strings based on an XML node's position in the document. For instance 'a3-e2-e' is the name produced for the 3rd attribute of the second element of the root element of the XML document. When producing the HTML for an XML node, I add an 'id' attribute to the HTML with the unique string of the XML node. Then in javascript when I execute an XPath I can discover the unique string of each node in the selected set and map each of them to their corresponding positions in the HTML.

With the hard part out of the way I changed the onload to get the fragment of the URI of the current document, interpret it as an XPath and highlight and navigate to the selected nodes. I also added an interactive floating bar from which you can enter your own XPaths and do the same. On a related note, I found that when accessing XML files via the file URI scheme the fragment is stripped off and not available to the javascript.

The next steps are of course to actually implement XPointer framework parsing as well as the limited number of schemes that the XPointer framework specifies.xml xpointer msxml res xpath xslt resource ie7 technical browser ie xsl

Resource Tools - IE7 XML Source View Upgrade Part 1

2007 May 9, 4:15I read about text/xml URI fragment resolution a few months ago. I was interested to find another kind of fragment reference other than the text/html URI fragment but of course I didn't find an implementation in IE, Firefox, or Opera. I decided to see how much work would be required to implement this in IE.

In IE and Firefox when you open an XML file that doesn't have an XML stylesheet the XML source is rendered with syntax highlighting. In IE I also noticed that the gold bar appears when you open an XML file off of your local machine. To me this suggested that the XML source was being rendered as HTML which I assumed was produced by running an XSLT on the source XML file. If so, I figured I could modify the XSLT to implement text/xml URI fragments. I ran FileMon to see if iexplore.exe loaded an XSLT file when opening an XML file. Only the XML file and MSXML3.DLL were opened and no XSLTs were loaded as files. My next hope for modifying the XSLT was if it existed as a resource in MSXML3.DLL. I did a findstr on the DLL for SCRIPT and found an XSLT so I decided to check for resources in MSXML3.DLL. Unfortunately my previous resource viewer didn't work correctly so I decided to write my own.

I created resource tools to view and modify resources in Windows modules. The viewer outputs HTML with links to the individual resources of a module using the res URI scheme that's built into IE. The modifier is a simple command line tool that replaces or adds one resource at a time to a module.

Using these tools I found that the XSLT was stored as a resource in MSXML3.DLL. I'll talk more about the existing XSLT and the one I replaced it with next time.resource technical xml msxml res xslt xsl