svglite does not play nice with xts

[paessens]

Hi dear Developers,
there seems to be a bug in svglite, hence, we cannot use the compiled svg-image without correction.

test case: [see test-svglite.R, files attached]

• import times series as xts
• specify filename (*.svg)
• plot, i.e. plot.xts
• first file is test2.svg and this cannot be opened in a viewer
• firefox shows error
  

Solution: the first 11 lines must be deleted!

Please could you investigate and remediate?
Many thanks and much appreciated.

Heinrich

Archive.zip

[paessens]

heck ... I forgot ...

[thomasp85]

This seems to be based on the CRAN version of svglite... The dev version includes a lot of rewrites, so can you check if the issue persists there?

[paessens]

yes, it's still there
Screenshot 2021-02-01 at 13.24.26.pdf

My function for a corrected svg-file is

Screenshot 2021-02-01 at 13.25.40.pdf

[paessens]

then I have a corrected svg like this

file2b551127afa2.svg.zip

[paessens]

sorry, I did this not in test

[thomasp85]

According to that you are still using the CRAN version of svglite?

[paessens]

well, this happened. which version is it?

[thomasp85]

You'll need to restart R to have the update take effect

[thomasp85]

Also, if it does persist please provide a reproducible example

[paessens]

OK, in test, I take the files from above Archive.zip

The CRAN version delivers file, but needs correction, as shown above.
The github version fails to write the file as requested by svglite('test2.svg') in the first step.

[thomasp85]

I cannot replicate this. Neither the error message nor any problems with the generated svg... Do you have write access to the location you are working in?

[paessens]

Yes i do

[thomasp85]

Is xts required to trigger the issue?

[paessens]

Yes, no issues with other data types crossed my path so far. I value svg and especially svglite very much since the file sizes can be kept small. That's also the reason why I use svg-buddy to optimize (even more). To keep typography consistent my preferred font (roboto condensed) is embedded, but only once for the dashboard (of 20 or more svg files).

[thomasp85]

Is there anything weird about your setup at all? Since I can't replicate any of this and since a graphic device have no concept of which packages are loaded (it only receives low-level instructions such as "draw a line there") this issue seems downright unsolvable unless we can somehow replicate it outside of your computer

[paessens]

Not that I know, everythng is fine considering setup on debian10 and macos11.15.7, on both machines the issue arises as you can see above. Let me check this issue on windows later the day.

[paessens]

here is testing windows -- same result, i.e. theres two def-tagged sections before the xml and svg tag that should not be there (the same applis to CRAN version)

[paessens]

here is the corrected svg file

[thomasp85]

Can you provide a reprex that does not rely on any loaded data

[paessens]

here you go -- same issue

[paessens]

[thomasp85]

Please use the reprex package as it ensures that your example is self contained and is not reliant on anything in your R setup. Have a look at https://reprex.tidyverse.org/index.html to familiarise yourself with the package

[paessens]

Sorry Thomas, I can't see any benefit from that. I consider the provided examples as simple, reproducible and self contained. The produced image should look like

and not like

[thomasp85]

I can see your issue now... In the future though, please do not use screen shots of code as it makes it extremely cumbersome for those you wish to help you reproduce your issue

[paessens]

here you go:

require(xts)
require(svglite)
sessionInfo()
# estr <- readRDS("estr.rds")
estr <- xts(1:10, Sys.Date()+1:10)
svglite('test2.svg')
plot(estr, main='€STR')
dev.off()

[paessens]

sorry I should know what a hashtag means in MD ;-)