Gephi forums

Re: Recent updates on GEXF time support

2010-12-08T21:38:51+01:00

Hi,

You do right, but this feature is currently not supported in Gephi.

Statistics:Posted by admin — 08 Dec 2010 21:38

Re: Recent updates on GEXF time support

2010-12-08T18:58:45+01:00

Hi,
i created a dynamic network with a dynamic attribute for nodes defined as:

CODE:

I would like for nodes to change their size according to this attribute. For that I am using in each node, for example:

CODE:

With this, nodes are properly shown (i.e. whenever corresponds according to the slices), however, their size don't change according to the attribute id="0" (which I set in Ranking > Nodes).

Any thoughts on what could be wrong? Or is this something not supported in Gephi 0.7 yet?

Thanks a lot.

Statistics:Posted by Luciano — 08 Dec 2010 18:58

Re: Recent updates on GEXF time support

2010-11-28T18:12:54+01:00

As noted by Skye Bender-deMoll, we don't use the same concept of slices as described by their paper and implemented in SoNIA.

I propose to use a more accurate word. "Spell" seems good as it is a moment or a period of time.

Statistics:Posted by admin — 28 Nov 2010 18:12

Re: Recent updates on GEXF time support

2010-11-17T18:45:10+01:00

It is possible to define it inside a GEXF file, but this is a missing feature in Gephi.

Statistics:Posted by admin — 17 Nov 2010 18:45

Re: Recent updates on GEXF time support

2010-10-31T19:39:00+01:00

Is it possible to dynamically define node size and colours according to time slices, as well as well as simply toggling the display or not of a node?

Statistics:Posted by psychemedia — 31 Oct 2010 19:39

Re: Recent updates on GEXF time support

2010-10-21T18:16:59+01:00

It's also important to let people define integers, just because it's a natural and efficient way to encode steps. Gephi can however handle these values as double internally.

Statistics:Posted by admin — 21 Oct 2010 18:16

Re: Recent updates on GEXF time support

2010-09-26T21:12:42+01:00

Yes I agree, using float was a mistake. Double is the default choice, as float is more narrow.

Statistics:Posted by mbastian — 26 Sep 2010 21:12

Re: Recent updates on GEXF time support

2010-09-26T21:13:58+01:00

I think this is preferable to use Double, since they have 64-bits floating point precision, while Float only have 32 bits.

This would allow developers to directly use timestamps, without need to convert it to a particular calendar reference, or XML format,
because timestamps are encoded in 64 bits too (if you still want to be able to load your GEXFs after 01-01-2038.. think about sustainable datasets )

Statistics:Posted by jbilcke — 26 Sep 2010 21:10

Re: Recent updates on GEXF time support

2010-09-26T20:35:24+01:00

I agree with all but the replacement of float by double:

I propose to replace float by double because double is how it is in Gephi.

Is not a good reason. What is the benefit of using doubles instead of floats?

Statistics:Posted by admin — 26 Sep 2010 20:35

[FIXED] Recent updates on GEXF time support

2010-09-26T18:02:38+01:00

Many work has been recently done for time support in the 0.7 beta, more detail coming soon. Some of these changes directly concern the GEXF format.

I propose the following changes/improvements:

Rename the "timetype" attribute in to "timeformat". This attribute is set on to specify how time information is encoded, either like a date or like a double.
The "timeformat" is currently either float or date and default value is date. I propose to replace float by double because double is how it is in Gephi. I also propose to set default value to double as is the one that is the less constrained, date needs a proper date (yyyy-mm-dd) formatting.
Add open intervals. Currently all intervals are inclusive. To be consistent we need open intervals as well. I propose therefore to add startopen and endopen attributes. Open intervals are also needed to convert fixed date points into time intervals concept. I explain

If I have a value at three different points in the time:
CODE:
```
2001  =>  2.12002  =>  5.22003  =>  6.5
```
converted to continuous, I believe we get:
CODE:
```
[2001, 2002)  =>  2.1[2002, 2003)  =>  5.2[2003, +inf]  =>  6.5
```

Another precision about dynamic attributes. In Gephi we now have new dynamic types for all static types (float, string, ...). I propose not to create new type keyword in GEXF but simply use these types when the type value is set to dynamic.

For example if I do:

CODE:

That would create DYNAMIC_STRING and DYNAMIC_FLOAT attribute columns in Gephi. As several attributes markup can be definied in a GEXF file, it's easy to mix static and dynamic types.

Statistics:Posted by mbastian — 26 Sep 2010 18:02