Spark Histogram are rendered as symbols in html/md

[ropensci]

Spark Histogram are rendered as symbols in html/md

[elinw]

Are you on Windows and did you read the instructions for Windows in the readme file?

[OmaymaS]

Yes I am on windows, I read the instructions, that's why I used Sys.setlocale("LC_CTYPE", "Chinese")

The point is, I see the spark histogram shown properly on the when I run the chunk in .Rmd file or when I View() the dataframe. But when I knit the file, the html/md show symbols.

Also on the console:

library(tidyverse)
library(skimr)

Sys.setlocale("LC_CTYPE", "Chinese")

> dat <- skim(storms) %>% filter(stat=="hist")

> dat
# A tibble: 10 x 5
           var    type  stat      level value
         <chr>   <chr> <chr>      <chr> <dbl>
 1        year numeric  hist ¨z¨z¨z¨}¨~¨}¨~¨~¨~¨}     0
 2       month numeric  hist ¨x¨x¨x¨x¨x¨y¨|¨~¨z¨x     0
 3         day integer  hist ¨~¨}¨}¨}¨}¨}¨}¨}¨}¨}     0
 4        hour numeric  hist ¨~¨x¨~¨x¨x¨~¨x¨~¨x¨x     0
 5         lat numeric  hist ¨y¨~¨~¨}¨~¨~¨|¨y¨x¨x     0
 6        long numeric  hist ¨x¨|¨~¨~¨~¨}¨}¨z¨x¨x     0
 7        wind integer  hist ¨y¨~¨|¨z¨y¨y¨x¨x¨x¨x     0
 8    pressure integer  hist ¨x¨x¨x¨x¨x¨x¨y¨z¨~¨y     0
 9 ts_diameter numeric  hist ¨~¨~¨|¨y¨x¨x¨x¨x¨x¨x     0
10 hu_diameter numeric  hist ¨~¨x¨x¨x¨x¨x¨x¨x¨x¨x     0

> dat$level[1]
[1] "▃▃▃▆▇▆▇▇▇▆"

[elinw]

Is this when knitting to PDF? I'll say that right now for me that is not working for hist at all (html is fine). In fact it stops processing. However adding latex_engine: xelatex to the YAML at least gets it to run. http://rmarkdown.rstudio.com/pdf_document_format.html#advanced_customization.

When did you install? Was it since this morning? There were changes made to the dependency that manages those, which I guess could have an impact.

[OmaymaS]

It happens with both, knitting to html as well pdf. I am more interested in the html now.
And I have installed it this week or 2 weeks ago, definitely before this morning.

[elinw]

Before I changed the latex engine pdf was dying. Now it knits but blank for both line_graph and hist. Html is working. I'm on a mac. I'll have to try to look into this but I suspect it might have to do with the character sets. I'm also going to see if pillar (formerly colformat) has this same issue outside of skimr.

[elinw]

I think the issue is that you need a font that supports the building block drawing characters an for braille. It seems like DejaVu would be the free one that is most commonly used. However, while I have figured out how to change the font of the text I haven't figured out how to get the <pre> regions to accept a new font. Also I haven't figured out how to access DejaVu.

[elinw]

@OmaymaS
Okay here is an update. The important thing, as described in the pillar readme file, is that you have a font installed that has block building characters. There are a number of these one of which is Dejavu Sans.

For my mac I used homebrew.

brew tap caskroom/fonts
brew cask install font-dejavu-sans

Later I found the extrafont package which includes several of the dejavu fonts.

At the top of my rmd file I changed my YAML to reference the font.

---
title: "Untitled"
output:
  pdf_document:
    latex_engine: xelatex
  html_document: default
font-family: "DejaVu Sans"
mainfont: "DejaVu Sans"
---

Remember that spaces matter in YAML so be careful.
HTML was working for me already and this continued fine. However now in PDF I still can't get the graphs embedded in the code chunk output that is formatted with <pre>. However, unlike previously I can get the spark graphs inline as below.

So I would suggest trying the extrafont package. If that doesn't work for Windows you will need to figure out how to install a font that supports block building and also braille.

If you could try that and then report back that would be awesome. There are a number of places that explain how depending on your version etc.

We still need to figure out the issue for pre formatted sections in PDFs.

[elinw]

Okay, one more update ... the real answer for me is using results=asis in the code chunk options and then using kable (or I assume something else) to format the table.

[OmaymaS]

Thanks!
I could finally knit to html by using kable. I added examples here. However, if I don't use kable, the histograms do not appear properly. Will check it out, check the PDF and get back to you.

[elinw]

@OmaymaS did you ever have an update on this? Now we have pull requests to support kable and pander. Can I close this issue and we can open a new one for a branch including that support? I think we also need to document some things about fonts. i think we may need to suggest extrafonts.

[koliajaykr]

@elinw
I'm sorry for late reply, PhD thesis writing torturing me. You are absolutely write about the types of font use and skimr summary output. For me, latex pdf bring histogram figure only with the fonts:
font-family: "DejaVu Sans"
mainfont: "DejaVu Sans", suggested by you. But the same file gives empty hist column if I use font:
mainfont: EB Garamond

I got this nice output with your present and earlier suggestions. Can you suggest something to remove the last 'value' column, it is full of zeros - seems not so important.

[elinw]

Yes since we have added the formatted values column the spark graphs have only the formats and are all set to 0. So in this case value is really not useful to display and could be dropped. That can be done by using dplyr::select() and dropping the value.

…

On Dec 2, 2017, at 11:22 PM, Ajay Kumar Koli ***@***.***> wrote: @elinw <https://github.com/elinw> I'm sorry for late reply, PhD thesis writing torturing me. You are absolutely write about the types of font use and skimr summary output. For me, latex pdf bring histogram figure only with the fonts: font-family: "DejaVu Sans" mainfont: "DejaVu Sans", suggested by you. But the same file gives empty hist column if I use font: mainfont: EB Garamond I got this nice output with your present and earlier suggestions. Can you suggest something to remove the last 'value' column, it is full of zeros - seems not so important. <https://user-images.githubusercontent.com/28312545/33522317-b0b0d21e-d80f-11e7-88a0-e675fd0c21e3.png> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#136 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAuEfRFsAUIys6DcWlqJbvFQ1pFJKKg7ks5s8iIRgaJpZM4PT1KW>.

[koliajaykr]

Thank you @elinw :)

[koliajaykr]

@elinw , I've a suggestion, is it not possible to provide the no. of observations and variables information too at footnote of the descriptive table? Because when we produce table using kable and skim, the info about the number of variables and observations is lost, the way it comes when we use only skim()

[elinw]

Oh that’s an interesting idea. I’m still not totally sure that we shouldn’t make some more flexible way of handling that data in general (so people could suppress it in regular printing). A footnote could make sense but so could a summary block (see the pull request for a summary). If you want to try to play around with that concept feel free to send a pull request.

…

On Dec 3, 2017, at 12:24 AM, Ajay Kumar Koli ***@***.***> wrote: @elinw <https://github.com/elinw> , I've a suggestion, is it not possible to provide the no. of observations and variables information too at footnote of the descriptive table? Because when we produce table using kable and skim, the info about the number of variables and observations is lost, the way it comes when we use only skim() — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#136 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAuEfQVAlU56F3_jFtlNYOYA-HXZZ9eLks5s8jCUgaJpZM4PT1KW>.

[koliajaykr]

@elinw , but I know exactly zero about pull request and how to contribute in it. :-(

[elinw]

I looked at the code and in pander() and kable() there is the caption parameter, so we can definitely put the separate table text in there at least for now. And adding the header text the way we do for print is also easy so we can do it for now (but long run I'd like something better). Pull request coming soon.

[koliajaykr]

Thank you @elinw :-)
Once, I'm done with my thesis. I'll make myself more friendly with the technicalities of GitHub, R packages and latex.

[elinw]

This is still an issue that we haven't solved. It's not clear if it is possible, but I'd rather at least have it showing as an issue so people know we are aware of it.

[koliajaykr]

Sure, you are right.

[elinw]

#209 Please test the develop branch if you can, it should resolve the Windows-histogram issue.