e4graph-tgraph — Technical discussions regarding tgraph, the Tcl binding for e4Graph

[E4graph-tgraph] 11'th Annual Tcl/Tk Conference

[<aku...@sh...>]

11'th Annual Tcl/Tk Conference

October 11 - 15, 2004
New Orleans, Louisiana, USA
Email Contact tc...@tc...

We are pleased to announce the 11'th Annual Tcl/Tk conference
(Tcl'2004), sponsored by Noumena Corporation, in cooperation with
ActiveState and ExpoTech.

Come to New Orleans to:

    * Learn about the power of Tcl/Tk.
    * Present exciting new work involving Tcl/Tk.
    * See the latest developments in Tcl/Tk.
    * Meet Tcl/Tk researchers and users from academia, government and industry.
    * Plan for future Tcl/Tk related developments.

    The conference program will include paper presentations,
    tutorials, Birds of a Feather (BOF) sessions and invited key-note
    talks. 

Registration

    Online registration is ready now.
    <http://www.tcl.tk/community/tcl2004/reg.html>

Tutorials

    Come learn about Tcl from the experts. This year's Tcl/Tk
    Conference includes one of the best sets of Tutorials ever offered
    including tutorials on Jacl, TclHttpd,
    Starkit, Advanced GUI construction, and the API.

    <http://www.tcl.tk/community/tcl2004/tut2004.html>

Schedule

    More details will be added to the schedule as they become
    available.

    <http://www.tcl.tk/community/tcl2004/schedule.html>

    Those attending the conference will be interested in the
    conference info page.

    <http://www.tcl.tk/community/tcl2004/info.html>

To keep in touch with news regarding the conference and Tcl events in
general, subscribe to the tcl-announce list.

<http://listserv.activestate.com/mailman/mysubs?show=announce>

Other Forms of Participation

    For those who are not presenting a paper at the conference, but
    would like to present their work in some form, we do provide
    several other forms of participation.

    Slots for Works-in-Progress (WIP) presentations and
    Birds-of-a-Feather sessions (BOFs) are available on a first-come,
    first-served basis by sending email to tc...@tc.... Some WIP
    and BOF time slots will be held open for on-site reservation, so
    we encourage all attendees with interesting work in progress to
    consider presenting that work at the conference.

Conference Committee

    Gerald Lester       HMS Software General Chair
    Andreas Kupries     ActiveState Corp
    Clif Flynt          Noumena Corp Website Admin
    Jeffrey Hobbs       ActiveState Corp
    Kevin Kenny         GE Global Research Center
    Ken Jones           Avia Training
    Mac Cody            Raytheon Company
    Kim Richerts  
    Steve Landers       Digital Smarties
    Sheila Miguez       Motorola
    Larry Virden        Tcl FAQ Maintainer

Contact Information
    
    tc...@tc...

[E4graph-tgraph] e4graph and tclkit--using the builtin metakit

[Brian T. <bt...@rr...>]

At
http://sourceforge.net/mailarchive/forum.php?thread_id=1924764&forum_id=6490,
Nicolas Boreto writes:

>I am now able to load tgraph and do some very preliminary things such as
>"tstorage open"...

I am also able to get e4graph to load into tclkit.  If I have e4graph.dll,
tgraph.dll (both from the e4graph binary download), and mk4vc60.dll
(downloaded from equi4.com) all in the same directory and execute load
tgraph.dll, then I am able to do some simple tgraph commands.  I haven't
gotten it to work as a starkit yet.  I think some special care is needed in
order to get the dlls that tgraph.dll depends on (e4graph.dll and
mk4vc60.dll) copied out of the vfs and into the real file system.

I wonder what it would take for tgraph.dll to use the metakit that is
already statically linked inside of tclkit?  Is there any way to do this?
It seems like a waste to have the metakit functionality loaded twice.  I
don't know much about the whole linking thing, but it seems to me like it
would require metakit to export some sort of stubs interface like the tcl
library does in order for it to work.  Is this correct, or is there
something simpler I'm missing?  I guess e4graph could be statically linked
into tclkit, but that is less desirable to me than the option of having
metakit loaded twice.

Brian Theado

[E4graph-tgraph] E4graph loads under Tclkit?

[Nicolas B. <nic...@ma...>]

Hello all,

This is my first post to this list as I've started looking at e4graph 
again.
Lat time I played with e4graph, I was not able to load it under tclkit, 
only under an installed tcl distro.
I am now able to load tgraph and do some very preliminary things such as 
"tstorage open"...

Are other using e4 under tclkit and is this approach reliable? My final 
intent would be to package my
app as a starkit/pack...

Also, do any tgraph users have any code snippets that they may share. I 
am in the process
of trying the samples in the docs...

thanx again for a cool package,

regards,

nicolas boretos
P.S Sorry if this gets posted twice, as I was not subscribed the first 
time I sent it...

[E4graph-tgraph] E4graph seems to load under Tclkit

[Nicolas B. <ni...@ma...>]

Hello all,

This is my first post to this list as I've started looking at e4graph again.
Lat time I played with e4graph, I was not able to load it under tclkit, 
only under an installed tcl distro.
I am now able to load tgraph and do some very preliminary things such as 
"tstorage open"...

Are other using e4 under tclkit and is this approach reliable? My final 
intent would be to package my
app as a starkit/pack...

Also, do any tgraph users have any code snippets that they may share. I 
am in the process
of trying the samples in the docs...

thanx again for a cool package,

regards,

nicolas boretos

Re: [E4graph-tgraph] tgraph2dot - render Tgraph structures to PNGs

[John E. <el...@lu...>]

Jacob Levy wrote:
> 
> OK, the mapping I'd use is as follows:
> 
> file                    storage
> graph                contents of storage
> node                  node
> node attr            vertex with type anything except node
> edge                   vertex with type node
> marker                edge from distinguished (implicit node) to any node
> edge attr            ????
> 
> A marker can mark only one node at a time. A marker can be moved from
> one node to another.
> I think that it works nicely to think of a marker as (in VizGraph terms)
> an edge from a special node to the designated node.

I can see why markers would be useful to a program for navigating
around a graph, but with your powerful iterators there wasn't any need
for them in extracting the graph.

I could render nodes that have markers in a different color perhaps?
But aren't markers more of a transient notion, like pointers?  What
would be the value of visualizing markers?
 
> I don't know what to do with GraphViz edge attributes when translating a
> GraphViz graph to an e4Graph storage. Hmmm..

I think the problem here is a mismatch in abstaction levels.  A graphviz
edge is more like a node with two parents in Tgraph. (except that it has
an existence dependency on both, not just on >0 )

So this means, I think, that a tool for visualizing Tgraphs would
work at a level where edges(vertices) have no attributes, but a tool for storing
dot graphs in Tgraphs would introduce a special node type for edges.

      dot
                edge
	node----------->node  


      Tgraph
                      edge
	node/vertex---node------vertex/node


Does this make sense?

Does Tgraph have a way to implement the existence dependancy of a node
on exactly two parent nodes?

John

Re: [E4graph-tgraph] tgraph2dot - render Tgraph structures to PNGs

[Jacob L. <jy...@be...>]

OK, the mapping I'd use is as follows:

file                    storage
graph                contents of storage
node                  node
node attr            vertex with type anything except node
edge                   vertex with type node
marker                edge from distinguished (implicit node) to any node
edge attr            ????

A marker can mark only one node at a time. A marker can be moved from 
one node to another.
I think that it works nicely to think of a marker as (in VizGraph terms) 
an edge from a special node to the designated node.

I don't know what to do with GraphViz edge attributes when translating a 
GraphViz graph to an e4Graph storage. Hmmm..

--JYL

John Ellson wrote:

>Jacob Levy wrote:
>
>>Wow, very nice! I don't have GraphViz installed on Win32, so I don't
>>*really* know what it looks like, but the code couldn't really be simpler.
>>
>>* You don't represent markers in the GraphViz graph. Any particular
>>reason? Isn't there an equivalent concept?
>>
>
>I don't yet fully understand the Tgraph concept of markers. My current
>understanding of the mapping is:
>
>	file            (= storage)
>	graph           (???)
>	node            (= node)
>	node-attributes (= vertex -type anything_but_node)
>	edge            (= vertex -type node)
>	edge-attributes (???)
>
>A graph is just a set of nodes and edges. Is that how I should think of markers?
>Can a marker have multiple nodes?
>
>>* Does GraphViz have an editor that knows how to handle large graphs?
>>
>
>There are three editors included in graphviz:
>	doted - written in Tcldot
>	dotty - written in lefty
>	dged  - written in Tcldgl/Tcldgr (an incremental layout engine)
>
>All can conveniently handle graphs up to a 100 or so visible nodes.
>
>Dot (and Tcldot) are able to render graphs up 1000 or so nodes.
>
>>If so, what about the other direction -- GraphViz --> Tgraph?
>>
>
>I've already made a start on this.   Its about the same complexity as tgraph2dot.
>
>There is no concept of type (other than string) in dot (or tcl) so
>type will have to be overloaded onto attribute names in some way.
>
>I also hacked an "xml2dot" which can produce graphs from html or xml.
>However these docs tend to be simple trees, so they are less interesting
>graphically.
>
>>Is it feasible to use Tgraph as a storage mechanism for GraphViz?
>>
>
>I think the translations are possible, so yes it could be used for
>an external storage mechanism.
>
>Using Tgraph as an in-memory structure manager would be more
>work with unknown benefits.
>
>>I'm working
>>on an editor for Tgraph, but so far I haven't come up with editors that
>>can support multiple parents (incoming edges).
>>
>
>i.e. directed graphs.   Yes, you have to solve the layout problem
>in order to provide a graphical editor for graphs that are more
>general than trees.
>
>>* I'm really pleased that there seems to be no need for any new C level
>>APIs to go from Tgraph to GraphViz. Cool! If you need more API for the
>>other direction let me know.
>>
>
>
>I could use your advice on the correct interpretation of markers,
>also ideas on how to map the graphviz notion of an edge-attribute?
>
>John Ellson
>

Re: [E4graph-tgraph] patch - package name fixes for Txml

[John E. <el...@lu...>]

Jacob Levy wrote:
> 
> OK, I've incorporated this one too. Can you explain what is wrong with
> calling Tgraph_Init? Are you worried that I'd put one-time code there
> and calling it multiple times would be wrong somehow?

Mostly its redundant once the Tcl_PkgRequire mechanism works properly.

Certainly if you had put one-time code in there it would be wrong to call
it twice.

John

Re: [E4graph-tgraph] tgraph2dot - render Tgraph structures to PNGs

[John E. <el...@lu...>]

Jacob Levy wrote:
> 
> Wow, very nice! I don't have GraphViz installed on Win32, so I don't
> *really* know what it looks like, but the code couldn't really be simpler.
> 
> * You don't represent markers in the GraphViz graph. Any particular
> reason? Isn't there an equivalent concept?

I don't yet fully understand the Tgraph concept of markers. My current
understanding of the mapping is:

	file            (= storage)
	graph           (???)
	node            (= node)
	node-attributes (= vertex -type anything_but_node)
	edge            (= vertex -type node)
	edge-attributes (???)

A graph is just a set of nodes and edges. Is that how I should think of markers?
Can a marker have multiple nodes?

> * Does GraphViz have an editor that knows how to handle large graphs?

There are three editors included in graphviz:
	doted - written in Tcldot
	dotty - written in lefty
	dged  - written in Tcldgl/Tcldgr (an incremental layout engine)

All can conveniently handle graphs up to a 100 or so visible nodes.

Dot (and Tcldot) are able to render graphs up 1000 or so nodes.

> If so, what about the other direction -- GraphViz --> Tgraph?

I've already made a start on this.   Its about the same complexity as tgraph2dot.

There is no concept of type (other than string) in dot (or tcl) so
type will have to be overloaded onto attribute names in some way.

I also hacked an "xml2dot" which can produce graphs from html or xml.
However these docs tend to be simple trees, so they are less interesting
graphically.

> Is it feasible to use Tgraph as a storage mechanism for GraphViz?

I think the translations are possible, so yes it could be used for
an external storage mechanism.

Using Tgraph as an in-memory structure manager would be more
work with unknown benefits.

> I'm working
> on an editor for Tgraph, but so far I haven't come up with editors that
> can support multiple parents (incoming edges).

i.e. directed graphs.   Yes, you have to solve the layout problem
in order to provide a graphical editor for graphs that are more
general than trees.

> * I'm really pleased that there seems to be no need for any new C level
> APIs to go from Tgraph to GraphViz. Cool! If you need more API for the
> other direction let me know.


I could use your advice on the correct interpretation of markers,
also ideas on how to map the graphviz notion of an edge-attribute?

John Ellson

Re: [E4graph-tgraph] patch - package name fixes for Txml

[Jacob L. <jy...@be...>]

OK, I've incorporated this one too. Can you explain what is wrong with 
calling Tgraph_Init? Are you worried that I'd put one-time code there 
and calling it multiple times would be wrong somehow?

Thanks! --JYL

John Ellson wrote:

>--- t4xml/src/t4xml.cpp.orig	Mon Aug 20 16:44:36 2001
>+++ t4xml/src/t4xml.cpp	Sun Aug 26 11:11:10 2001
>@@ -744,17 +744,10 @@
> T4XML_RealInit(Tcl_Interp *interp)
> {
>     /*
>-     * If we got this far, we must have loaded libtgraph. Ensure the
>-     * TGraph package is registered, by calling the init function.
>-     */
>-
>-    Tgraph_Init(interp);
>-
>-    /*
>      * Require the T4Graph package.
>      */
> 
>-    if (Tcl_PkgRequire(interp, "TGraph", (char *) "1.0.0", 0) == NULL) {
>+    if (Tcl_PkgRequire(interp, "Tgraph", (char *) "1.0.0", 0) == NULL) {
> 	return TCL_ERROR;
>     }
> 
>@@ -775,5 +768,5 @@
>      * All went well, so declare that we provide the T4XML package.
>      */
> 
>-    return Tcl_PkgProvide(interp, (char *) "TXML", (char *) "1.0.0");
>+    return Tcl_PkgProvide(interp, (char *) "Txml", (char *) "1.0.0");
> }
>
>_______________________________________________
>E4graph-tgraph mailing list
>E4g...@li...
>http://lists.sourceforge.net/lists/listinfo/e4graph-tgraph
>

Re: [E4graph-tgraph] tgraph2dot - render Tgraph structures to PNGs

[Jacob L. <jy...@be...>]

Wow, very nice! I don't have GraphViz installed on Win32, so I don't 
*really* know what it looks like, but the code couldn't really be simpler.

* You don't represent markers in the GraphViz graph. Any particular 
reason? Isn't there an equivalent concept?

* Does GraphViz have an editor that knows how to handle large graphs? If 
so, what about the other direction -- GraphViz --> Tgraph? Is it 
feasible to use Tgraph as a storage mechanism for GraphViz? I'm working 
on an editor for Tgraph, but so far I haven't come up with editors that 
can support multiple parents (incoming edges).

* I'm really pleased that there seems to be no need for any new C level 
APIs to go from Tgraph to GraphViz. Cool! If you need more API for the 
other direction let me know.

--JYL

John Ellson wrote:

>Jacob Levy wrote:
>
>>Connecting e4Graph and GraphViz sounds like a really fun project! Any
>>way I can help, just ask!
>>
>
>OK. Try this on one of your Tgraph databases and see if the result
>is potentially useful.
>
>John Ellson
>
>
>------------------------------------------------------------------------
>
>#!/usr/bin/tclsh
>#
># tgraph2dot - convert a Tgraph to a Tcldot graph and render 
>#			to PNG and DOT files
>#
># John Ellson <el...@lu...>
>#
># requires e4graph:   http://e4graph.sourceforge.net/
>#          graphviz:  http://www.graphviz.org/
>#
># usage:  tgraph2dot foo.db
>#         ==> foo.db.png and foo.db.dot
>
># Tgraph is in a non-default location
>lappend auto_path /usr/local/lib
>package require Tgraph
>
># Tcldot was installed in the default location from graphviz rpms
>package require Tcldot
>
># convert the Tgraph from the database file specified
>set f [lindex $argv 0]
>if {! [file readable $f]} {
>	puts stderr "file \"$f\" is missing or not readable"
>	exit
>}
>set s [tgraph::open $f]
>
># create a new Tcldot graph structure for the conversion
>set g [dotnew digraph]
>
># create all nodes (even if they have no edges or attributes)
>$s foreach node n {
>	# use the Tgraph namespace name of the node as its name in Tcldot
>	$g addnode $n label [lindex [split $n :] end]
>}
>
># create all edges and attributes
>$s foreach vertex v {
>	# if the type is "node" then it is an edge, else it is an attribute
>	if {[string equal [$v type] node]} {
>		$g addedge [$v node] [$v get] label [$v name]
>	} {
>		# the node should already exists, this just adds the
>		#     attribute name and value to the node
>		$g addnode [$v node] [$v name] [$v get]
>	}
>}
>
># generate a PNG image
>set fpng [open $f.png w]
>$g write $fpng PNG
>close $fpng
>
># generate canonical dot version of graph
>set fdot [open $f.dot w]
>$g write $fdot CANON
>close $fdot
>
>#clean up
>$g delete
>$s close
>
> tgraph2dot
>
> Content-Type:
>
> text/plain
> Content-Encoding:
>
> 7bit
>
>

[E4graph-tgraph] tgraph2dot - render Tgraph structures to PNGs

[John E. <el...@lu...>]

Jacob Levy wrote:
> 
> Connecting e4Graph and GraphViz sounds like a really fun project! Any
> way I can help, just ask!

OK. Try this on one of your Tgraph databases and see if the result
is potentially useful.

John Ellson

Re: [E4graph-tgraph] pkgIndex.tcl ?

[Jacob L. <jy...@be...>]

Sorry, couldn't parse that -- it doesn't *what*? Doesn't work, presumably?

Anyways,  the default install location for Unix is already 
/usr/local/lib which presumably should be where the user is already 
looking, and on Win32 I intend (with an installer) to put the DLLs in 
c:\windows or equivalent. So this shouldn't be that much of an issue.

--JYL

John Ellson wrote:

>Jacob Levy wrote:
>
>>One reason I didn't yet create a package index is because libtgraph.so
>>depends on libe4graph.so and libmk4.so on Unix, and similarly on DLLs by
>>that name on Win32. If the directory containing these depended-on
>>libraries is not in the LD_LIBRARY_PATH (on Unix) or path (on Win32)
>>then I'm not sure a 'package require tgraph' which would attempt to load
>>libtgraph.so would work. Do you have any insights into that from your
>>experience with GraphViz?
>>
>
>It doesn't.
>
>Personally I don't consider the use of LD_LIBRARY_PATH to be an
>acceptable
>solution to this problem since setting it is too much to expect of the
>targeted modern Linux Desktop user. I much prefer RedHat's solution of
>using rpm to manage packages well enough to put everything into
>/usr/lib.
>Then, so long as /sbin/ldconfig has been run post-install, library
>dependencies are properly resolved.
>
>On linux you can extend the paths seached for libraries by adding, e.g.,
>/usr/local/lib to /etc/ld.so.conf.  This avoids the use of
>LD_LIBRARY_PATH
>for development packages on Linux, but I never did find the equivalent
>of
>/etc/ld.so.conf on Solaris.  Would you know?
>
>GNUs libtool really helped me with these cross-platform shared-library
>complexities. Highly recommended.
>
>>Connecting e4Graph and GraphViz sounds like a really fun project! Any
>>way I can help, just ask!
>>
>
>OK, thanks.  I need to play a bit more with e4graph to understand what
>it does, then I'm sure I'll be back.
>
>John
>

Re: [E4graph-tgraph] patch - package name must match entry point

[Jacob L. <jy...@be...>]

Thanks! It's in the source base now and will be in b1.

--JYL

John Ellson wrote:

>In order for tcl's "package require" to work
>the package name must match the Tgraph_Init 
>entry point name. 
>
>-------------------------
>
>--- tcl/src/t4init.cpp.orig	Sun Aug 26 10:38:44 2001
>+++ tcl/src/t4init.cpp	Sun Aug 26 10:41:45 2001
>@@ -116,7 +116,7 @@
>      * All went well, so declare that we provide the T4Graph package.
>      */
> 
>-    return Tcl_PkgProvide(interp, (char *) "TGraph", (char *) "1.0.0");
>+    return Tcl_PkgProvide(interp, (char *) "Tgraph", (char *) "1.0.0");
> }
> 
> /*
>
>_______________________________________________
>E4graph-tgraph mailing list
>E4g...@li...
>http://lists.sourceforge.net/lists/listinfo/e4graph-tgraph
>

[E4graph-tgraph] patch - package name fixes for Txml

[John E. <el...@lu...>]

--- t4xml/src/t4xml.cpp.orig	Mon Aug 20 16:44:36 2001
+++ t4xml/src/t4xml.cpp	Sun Aug 26 11:11:10 2001
@@ -744,17 +744,10 @@
 T4XML_RealInit(Tcl_Interp *interp)
 {
     /*
-     * If we got this far, we must have loaded libtgraph. Ensure the
-     * TGraph package is registered, by calling the init function.
-     */
-
-    Tgraph_Init(interp);
-
-    /*
      * Require the T4Graph package.
      */
 
-    if (Tcl_PkgRequire(interp, "TGraph", (char *) "1.0.0", 0) == NULL) {
+    if (Tcl_PkgRequire(interp, "Tgraph", (char *) "1.0.0", 0) == NULL) {
 	return TCL_ERROR;
     }
 
@@ -775,5 +768,5 @@
      * All went well, so declare that we provide the T4XML package.
      */
 
-    return Tcl_PkgProvide(interp, (char *) "TXML", (char *) "1.0.0");
+    return Tcl_PkgProvide(interp, (char *) "Txml", (char *) "1.0.0");
 }

[E4graph-tgraph] patch - package name must match entry point

[John E. <el...@lu...>]

In order for tcl's "package require" to work
the package name must match the Tgraph_Init 
entry point name. 

-------------------------

--- tcl/src/t4init.cpp.orig	Sun Aug 26 10:38:44 2001
+++ tcl/src/t4init.cpp	Sun Aug 26 10:41:45 2001
@@ -116,7 +116,7 @@
      * All went well, so declare that we provide the T4Graph package.
      */
 
-    return Tcl_PkgProvide(interp, (char *) "TGraph", (char *) "1.0.0");
+    return Tcl_PkgProvide(interp, (char *) "Tgraph", (char *) "1.0.0");
 }
 
 /*

Re: [E4graph-tgraph] pkgIndex.tcl ?

[John E. <el...@lu...>]

Jacob Levy wrote:
> 
> One reason I didn't yet create a package index is because libtgraph.so
> depends on libe4graph.so and libmk4.so on Unix, and similarly on DLLs by
> that name on Win32. If the directory containing these depended-on
> libraries is not in the LD_LIBRARY_PATH (on Unix) or path (on Win32)
> then I'm not sure a 'package require tgraph' which would attempt to load
> libtgraph.so would work. Do you have any insights into that from your
> experience with GraphViz?

It doesn't.

Personally I don't consider the use of LD_LIBRARY_PATH to be an
acceptable
solution to this problem since setting it is too much to expect of the
targeted modern Linux Desktop user. I much prefer RedHat's solution of
using rpm to manage packages well enough to put everything into
/usr/lib.
Then, so long as /sbin/ldconfig has been run post-install, library
dependencies are properly resolved.

On linux you can extend the paths seached for libraries by adding, e.g.,
/usr/local/lib to /etc/ld.so.conf.  This avoids the use of
LD_LIBRARY_PATH
for development packages on Linux, but I never did find the equivalent
of
/etc/ld.so.conf on Solaris.  Would you know?

GNUs libtool really helped me with these cross-platform shared-library
complexities. Highly recommended.

> Connecting e4Graph and GraphViz sounds like a really fun project! Any
> way I can help, just ask!

OK, thanks.  I need to play a bit more with e4graph to understand what
it does, then I'm sure I'll be back.

John

Re: [E4graph-tgraph] pkgIndex.tcl ?

[Jacob L. <jy...@be...>]

One reason I didn't yet create a package index is because libtgraph.so 
depends on libe4graph.so and libmk4.so on Unix, and similarly on DLLs by 
that name on Win32. If the directory containing these depended-on 
libraries is not in the LD_LIBRARY_PATH (on Unix) or path (on Win32) 
then I'm not sure a 'package require tgraph' which would attempt to load 
libtgraph.so would work. Do you have any insights into that from your 
experience with GraphViz?

Connecting e4Graph and GraphViz sounds like a really fun project! Any 
way I can help, just ask!

--JYL

John Ellson wrote:

>Jacob Levy wrote:
>
>>Unfortunately I didn't create one. Can you create it and email it to me?
>>Then I'll respin the source release.
>>
>>Thanks! --JYL
>>
>
>One problem is that you can't create one in /usr/local/lib because 
>it might already exist for other packages.
>
>The approach I used in graphviz was to put all the dlls and the
>pkgIndex.tcl
>into a graphviz subdirectory under <prefix>/lib/   The tcl package
>mechanism
>automatically searches all subdirectories so it can find your private 
>pkgIndex.tcl in there.  
> 
>You are already installing the dlls in <prefix>/lib rather
>than <prefix>/lib/<package>, but thats ok, you can still just
>put a pkgIndex.tcl in a subdirectory.
>
>So something like:
>
>	/usr/local/lib/e4graph/pkgIndex.tcl
>
>should contain something like:
>
>	package ifneeded TGraph 1.0 "
>        	load [file join $dir .. libtgraph.so] TGraph"
>	package ifneeded Txml 1.0 "
>        	load [file join $dir .. libtxml.so] Txml"
>
>
>You might also want to do a "/sbin/ldconfig" in "make install"
>and also warn the user that if tclsh is in /usr, but this package is in 
>/usr/local, that the search path will have to be extended, eg.
>by "lappend auto_path /usr/local/lib"  before the "package require"
>
>Hope this helps.
>
>BTW, do you have some example graphs in e4graph?   I was
>going to see what was involved in connecting e4graph to the
>graph layout engines in graphviz.  See: http://www.graphviz.org/ 
>
>John
>
>_______________________________________________
>E4graph-tgraph mailing list
>E4g...@li...
>http://lists.sourceforge.net/lists/listinfo/e4graph-tgraph
>

Re: [E4graph-tgraph] pkgIndex.tcl ?

[John E. <el...@lu...>]

Jacob Levy wrote:
> 
> Unfortunately I didn't create one. Can you create it and email it to me?
> Then I'll respin the source release.
> 
> Thanks! --JYL

One problem is that you can't create one in /usr/local/lib because 
it might already exist for other packages.

The approach I used in graphviz was to put all the dlls and the
pkgIndex.tcl
into a graphviz subdirectory under <prefix>/lib/   The tcl package
mechanism
automatically searches all subdirectories so it can find your private 
pkgIndex.tcl in there.  
 
You are already installing the dlls in <prefix>/lib rather
than <prefix>/lib/<package>, but thats ok, you can still just
put a pkgIndex.tcl in a subdirectory.

So something like:

	/usr/local/lib/e4graph/pkgIndex.tcl

should contain something like:

	package ifneeded TGraph 1.0 "
        	load [file join $dir .. libtgraph.so] TGraph"
	package ifneeded Txml 1.0 "
        	load [file join $dir .. libtxml.so] Txml"


You might also want to do a "/sbin/ldconfig" in "make install"
and also warn the user that if tclsh is in /usr, but this package is in 
/usr/local, that the search path will have to be extended, eg.
by "lappend auto_path /usr/local/lib"  before the "package require"

Hope this helps.

BTW, do you have some example graphs in e4graph?   I was
going to see what was involved in connecting e4graph to the
graph layout engines in graphviz.  See: http://www.graphviz.org/ 

John

Re: [E4graph-tgraph] pkgIndex.tcl ?

[Jacob L. <jy...@be...>]

Unfortunately I didn't create one. Can you create it and email it to me? 
Then I'll respin the source release.

Thanks! --JYL

John Ellson wrote:

>I did:
>	./configure --enable-tcl --enable-xml
>	make 
>	make install
>
>and everything seemed to go ok, but
>	package require TGraph
>isn't working.
>
>I didn't see a pkgIndex.tcl installed for this package.  Did I
>miss a step somewhere?
>
>John Ellson
>
>_______________________________________________
>E4graph-tgraph mailing list
>E4g...@li...
>http://lists.sourceforge.net/lists/listinfo/e4graph-tgraph
>

[E4graph-tgraph] pkgIndex.tcl ?

[John E. <el...@lu...>]

I did:
	./configure --enable-tcl --enable-xml
	make 
	make install

and everything seemed to go ok, but
	package require TGraph
isn't working.

I didn't see a pkgIndex.tcl installed for this package.  Did I
miss a step somewhere?

John Ellson

[E4graph-tgraph] RE: tgraph usage

[Horst E. <ho...@fr...>]

Hello Jacob,
as I said, I am slowly finding my way around tgaph.

So this time I do have a performance question:

my data is in a structure like:

....
<entry>
  <form>
    <orth>this word</orth>
  </form>
  <trans>translation of word</trans>
</entry>
<entry>
 ... next entry 
</entry>
... many other entries (approx 100'000) ...

I would like to select a "entry" in accordance to the word given in the "orth"
element. I tried a foreach loop on all the entries ( ParentNode foreach -name
"entry" ...) in the foreach loop I select the "form" and "orth" vertex. After
about 50'000 my memory is used up, swap increases and everything slows down
very much :(
I was thinking about adding a attribute to each entry with the relevant word. I
am reluctant to do that as inconsistency issues might arise.

Thanks,

Horst


On 05-Mar-01 jacob levy wrote:
> Hi Horst
> 
> Thanks for trying e4Graph, and thanks for your note, I am the author of
> e4Graph (my day job is at Sun Micro :). I'll respond in more detail
> later, just a short note to let you know I received your email.
> 

Ho...@fr...
Horst Eyermann 
Germany

You need a dictionary? - visit http://www.freedict.de
for free (GPL) dictionaries (unix; windows work in progress)
For windows, visit http://www.freedict.de/wbuch

A article (in German) about dictionary efforts on the net
http://www.heise.de/tp/deutsch/inhalt/on/5927/1.html 

[E4graph-tgraph] how to use tgraph

[Horst E. <ho...@fr...>]

Hello,
I tried to use e4graph (txml) in order to edit a XML Document (a
bilingual dictionary).
However I had problems with the documentation, so my hope is that someone on
this list can help me.

I could generate a new database, and set markers and nodes without problems.
There were also no problems reading a xml-document.

But I am not sure on how to work with a document I already generated. 

That is after I opened the database how do I find the root node?

If someone could send me some example code, or a pointer to more documentation
it would be very great.

Did someone write some interface similar to XPath in order to find a specific
node? - For my project this would be very helpful. Alternative I would like to
know if anyone has ideas on what to do / not to do.

Regards,

Horst



Ho...@fr...
Horst Eyermann 
Germany

You need a dictionary? - visit http://www.freedict.de
for free (GPL) dictionaries (unix; windows work in progress)
For windows, visit http://www.freedict.de/wbuch

A article (in German) about dictionary efforts on the net
http://www.heise.de/tp/deutsch/inhalt/on/5927/1.html 

Re: [E4graph-tgraph] Re: Running tgraph

[Jacob L. <jy...@be...>]

Yeah the "cmd" needs to be one token. Rewrite as:

    $s foreach marker mymarker {puts "[$mymarker name]"}

--JYL

John Fletcher wrote:

> Date sent:              Wed, 01 Nov 2000 20:53:37 -0800
> From:                   Jacob Levy <jy...@be...>
> Organization:           Leage of Retired Slug Programmers
> To:                     John Fletcher <J.P...@as...>
> Subject:                Re: [E4graph-tgraph] Re: Running tgraph
>
> > I misspoke. It turns out that Alpha 1 already contains some of these facilities. Here's
> > what's implemented already in the version you already have installed:
> >
> > For storages:
> >
> > $s foreach marker m body-code
> > $s foreach node n body-code
> > $s foreach vertex v body-code
> >
>
> Jacob
>
> I am testing the above and the grammar seems different.
>
> When I do this, where mystorage is a valid storage and mymarker
> a valid marker,
>
> % $mystorage foreach marker $mymarker \
> puts " [ $mymarker name ] "
>
> (Note: line broken at \)
>
> I get:
>
> wrong # args: should be " $storage foreach marker var ?-keep
> bool? cmd"
>
> What is up please?  Have you a working example?  How could I
> declare an object of marker for this if I didn't know a name?
>
> John

Re: [E4graph-tgraph] Re: Running tgraph

[John F. <J.P...@as...>]

Date sent:      	Wed, 01 Nov 2000 20:53:37 -0800
From:           	Jacob Levy <jy...@be...>
Organization:   	Leage of Retired Slug Programmers
To:             	John Fletcher <J.P...@as...>
Subject:        	Re: [E4graph-tgraph] Re: Running tgraph

> I misspoke. It turns out that Alpha 1 already contains some of these facilities. Here's
> what's implemented already in the version you already have installed:
> 
> For storages:
> 
> $s foreach marker m body-code
> $s foreach node n body-code
> $s foreach vertex v body-code
> 

Jacob

I am testing the above and the grammar seems different.

When I do this, where mystorage is a valid storage and mymarker 
a valid marker, 

% $mystorage foreach marker $mymarker \
puts " [ $mymarker name ] "

(Note: line broken at \)

I get: 

wrong # args: should be " $storage foreach marker var ?-keep 
bool? cmd"


What is up please?  Have you a working example?  How could I 
declare an object of marker for this if I didn't know a name?

John

Re: [E4graph-tgraph] Re: Running tgraph

[John F. <J.P...@as...>]

Date sent:      	Wed, 01 Nov 2000 20:53:37 -0800
From:           	Jacob Levy <jy...@be...>
Organization:   	Leage of Retired Slug Programmers
To:             	John Fletcher <J.P...@as...>
Subject:        	Re: [E4graph-tgraph] Re: Running tgraph

> I misspoke. It turns out that Alpha 1 already contains some of these facilities. Here's
> what's implemented already in the version you already have installed:
> 
> For storages:
> 
> $s foreach marker m body-code
> $s foreach node n body-code
> $s foreach vertex v body-code
> 
> For nodes:
> 
> $n foreach vertex v ?-type t? ?-name n? body-code
> 
> Not implemented:
> 
> $n foreach parent p body-code
> $n foreach marker m body-code
> 
> Sorry about the confusion.
> 
> --JYL
> 
Jacob, 

O.K., I'll check it out.

John

-------------------------------------------------------------------
Dr John P. Fletcher          Tel: (44) 121 359 3611 ext 4625
Chemical Engineering and Applied Chemistry (CEAC),
School of Engineering and Applied Science (SEAS),
Aston University,            Fax: (44) 121 359 4094
Aston Triangle,              Email: J.P...@as...
BIRMINGHAM B4 7ET  U.K.      CEAC Web site http://www.ceac.aston.ac.uk/