get_citation_history(id) gives error for some id's

[wdarling]

Hi. For some reason when I try to get my citation history I get the following error:
Error in data.frame(year = years, cites = vals) :
arguments imply differing number of rows: 9, 8
This doesn't happen for many other IDs that I've tried. My id is 'xpECwJQAAAAJ'... do you know why it doesn't work? Too few citations? ;-)

[jefferis]

Too few citations? ;-)

Sort of. The code relies on finding years and bars indicating number of citations separately. You have one year (2009) in your chart without citations. There is no 0 height bar for 2009. Looking at the code I don't see an easy fix without reworking the code e.g. to look at the x axis position of the bars and figure out where there are gaps.

[jefferis]

Actually there seems some information in the style tag. This fixes for me (see if(length(years)>length(vals)){):

get_citation_history <- function(id) {

    ## Ensure only one ID
    id <- tidy_id(id)

    ## Read the page and parse the key data
    url_template <- "http://scholar.google.com/citations?hl=en&user=%s&pagesize=100&view_op=list_works"
    url <- sprintf(url_template, id)

    ## A better way would actually be to read out the plot of citations
    page <- GET(url, handle=getOption("scholar_handle")) %>% read_html()
    years <- page %>% html_nodes(xpath="//*/span[@class='gsc_g_t']") %>%
        html_text() %>% as.numeric()
    vals <- page %>% html_nodes(xpath="//*/span[@class='gsc_g_al']") %>%
        html_text() %>% as.numeric()
    if(length(years)>length(vals)){
      # Some years don't have citations. 
      # We need to match the citation counts and years
      # <a href="javascript:void(0)" class="gsc_g_a" style="left:8px;height:5px;z-index:9">\n  <span class="gsc_g_al">2</span>\n</a>
      style_tags=page %>% html_nodes(css = '.gsc_g_a') %>% 
        html_attr('style')
      # these z indices seem to be the indices starting with the last year
      zindices=as.integer(stringr::str_match(style_tags, 'z-index:([0-9]+)')[,2])
      # empty vector of 0s
      allvals=integer(length=length(years))
      # fill in 
      allvals[zindices]=rev(vals)
      # and then reverse
      vals=rev(allvals)
    }
    df <- data.frame(year=years, cites=vals)

    return(df)
}

[wdarling]

Yes that did the trick! Thanks 👍

[wdarling]

Hi, unfortunately I spoke too soon. The fix makes the function not crash, but it doesn't get the years and citations lined up properly. The problem is these two lines:

# fill in
allvals[zindices]=rev(vals)
# and then reverse
vals=rev(allvals)

When you do a reverse, the missing value then lines up with the second most recent year instead of the second oldest year. Here is some code that works (though you might have a more efficient way of doing it):

zindices=as.integer(stringr::str_match(style_tags, 'z-index:([0-9]+)')[,2])
allvals=integer(length=length(years))
i = 1
j = 1
prev = length(years)+1
for(z in zindices) {
  if((z+1) != prev) {
    i = i+1
  }
  allvals[i] = vals[j]    
  i = i+1
  j = j+1
  prev = z
}

Basically what I'm doing is checking if the previous zindex has a break in continuity. If it does then there were no citations for this year and we should just use the 0 that was already placed in allvals.

[jefferis]

Sorry you're right there was one rev too many. It should have been:

      # fill in 
      allvals[zindices]=vals
      # and then reverse
      vals=rev(allvals)

[jefferis]

Try devtools::install_github("jefferis/scholar@develop")

[wdarling]

Ok -- this time it works 😄 Thanks