text labels on x-y scatterplots

[Christian Ritter <christian ritter ridaco be>]

Hi gnumericists,

I've been watching gnumeric grow and every once in a while I try it to 
see if I can already use it for my work. Congratulations, it's getting 
closer, but there remain some show stoppers.

I'm working under Windows (Vista) and occasionally under Debian Linux. 
The latest version I tried is 1.9.0 under Windows Vista. Here are some 
of the things which (still) keep me from abandoning Excel for some of my 
work:

1) I cannot label points on x-y graphs nor identify individual points in 
an easy way. This is something that Excel is not very good at, but at 
least one can do it. Ideally, it should be possible to create an x-y 
graph with a label, a color, and a symbol vector alongside, such that I 
can modify displayed text, color, and symbol via auxiliary columns in 
the data frame. This is something Excel can't do either, but one can 
force it (using the XYChartLabeler add-in). If gnumeric could do this, 
it would be a step ahead in another critical area (besides mathematical 
and statistical functions).

2) I cannot link gnumeric to the R language. Some years ago, Duncan 
Temple Lang wrote Rgnumeric, an experimental version to combine gnumeric 
and R in a linux environment. Yet, this package was not developed 
further (maybe gnumeric was too immature at that point in time) and has 
fallen into disrepair. On windows it does not work at all. This would 
also be a way out of graphics limitation: If I can have worksheet 
functions in gnumeric which invoke R graphics working on ranges in 
gnumerics then all the graphics power in R becomes available in 
gnumeric. For those who are interested on what this could look like, try 
RExcel (E Neuwirth and T Baier). There should be an R-package on cran 
which installs what is needed under windows within the next days.

3) Pivot tables are in the announced list but haven't come in yet ... 
again this could be circumvented via R using the reshape package.

4) I cannot access databases (ODBC, JDBC, or DBI) using an import 
mechanism from gnumeric. (Here is another thing one could do with 
gnumeric which Excel can't do: save a range as a table in a relational 
data base. Obviously, this only works if the range satisfies a few rules 
... but this can be done. Again, RExcel shows a way on how to 
automatically convert a range to an R dataframe. And from there it's 
only one call to a database... and maybe one doesn't have to go through 
R at all).

Any comments? Anyone to second my wishes? I'll be at the useR conference 
at Dortmund in summer and I'll try to find some souls interested in 
revisiting the Rgnumeric project ... I'm not much of a programmer myself 
... but I can help with ideas.

Chris

Christian Ritter
Ritter and Danielson Consulting
Brussels, Belgium.