ropensci-archive / rplos Goto Github PK

Hi - tried installing from cran (sudo R; install.packages('rplos')), got this:

[...]
installing to /usr/local/lib/R/site-library/RCurl/libs
** R
** data
** inst
** preparing package for lazy loading
Creating a new generic function for "close" in "RCurl"
** help
*** installing help indices
** building package indices ...
** testing if installed package can be loaded

* DONE (RCurl)
* installing *source* package ‘reshape’ ...
** R
** data
**  moving datasets to lazyload DB
** inst
** preparing package for lazy loading
** help
*** installing help indices
** building package indices ...
** testing if installed package can be loaded

* DONE (reshape)
* installing *source* package ‘googleVis’ ...
** R
** data
**  moving datasets to lazyload DB
** demo
** inst
** preparing package for lazy loading
** help
*** installing help indices
** building package indices ...
** testing if installed package can be loaded

* DONE (googleVis)
ERROR: dependency ‘stringr’ is not available for package ‘httr’
* removing ‘/usr/local/lib/R/site-library/httr’
ERROR: dependencies ‘ggplot2’, ‘stringr’, ‘httr’ are not available for package ‘rplos’
* removing ‘/usr/local/lib/R/site-library/rplos’

The downloaded packages are in
    ‘/tmp/RtmpInY9HC/downloaded_packages’
Warning messages:
1: In install.packages("rplos") :
  installation of package 'httr' had non-zero exit status
2: In install.packages("rplos") :
  installation of package 'rplos' had non-zero exit status

Any clues please?

Martin suggested "summary" as the default output, but I disagree:

We could have "summary" as the default option, but it seems like the summary option doesn't give you any metrics at all, just metadata. Perhaps it would make more sense to by default return the summary metrics for each data provider. I think this is returned together with the very detailed by day metrics within the "detail" option output. For example:

> out <- almplosallviews(doi='10.1371/journal.pone.0029797', info='detail')
> out[["metrics"]] # get metrics summary data.frame
                .id shares total citations  pdf  html comments likes groups
1         citeulike      1     1        NA   NA    NA       NA    NA     NA
2          crossref     NA     2         2   NA    NA       NA    NA     NA
3            nature     NA     3         3   NA    NA       NA    NA     NA
4            pubmed     NA     0         0   NA    NA       NA    NA     NA
5            scopus     NA     1         1   NA    NA       NA    NA     NA
6           counter     NA 23689        NA 2038 21553       NA    NA     NA
7  researchblogging     NA     1         1   NA    NA       NA    NA     NA
8              biod     NA    16        NA    0     0       NA    NA     NA
9               wos     NA     1         1   NA    NA       NA    NA     NA
10              pmc     NA   173        NA   33   140       NA    NA     NA
11         facebook    133   176        NA   NA    NA       18    25     NA
12         mendeley     29    30        NA   NA    NA       NA    NA      1
13          twitter     NA     5        NA   NA    NA        5    NA     NA
14        wikipedia    169   208        39   NA    NA       NA    NA     NA

Martin said something about copyright and author roles fields as not documented or not in the output of fxn calls. look in to this.

From this thread on SO http://stackoverflow.com/questions/22266705/how-to-search-pubmed-or-other-databases-using-r/22271647

-Making progress, writing the default search function now.
-Need to figure out a compact way of displaying results, and the best way of doing so.
-Include some graphics functions summarizing results.
-Can datasets be downloaded in addition to just the bibliographic information?

Can filter by altmetrics the search results got back...

See here: http://blogs.plos.org/tech/advanced-search-api-capabilties/

http://www.plosone.org/article/fetchObject.action?uri=info%3Adoi%2F10.1371%2Fjournal.pone.0086169&representation=XML

Just append the DOI to the base url. and add the representation parameter

Direct quote from user's email:

I also had another question, and possible bug report. we've been trying to search all PLoS papers ever for data, using the code below.

f <- "id, results_and_discussion, journal, cross_published_journal_key, subject_level_1, publication_date"

num.results <- 250 # how many papers per search. I think it is best to keep this limited, to reduce strain on memory and the PLoS API. 250 seems fine.

for(start in seq(1, 300000, num.results)){

    print(paste("Starting from record", start))

    # get a list of 250 results (note that we don't want to keep these around)
    x <- searchplos("*:*", fields=f, toquery="doc_type:full", returndf = FALSE, start=start, limit=num.results) 

    # sleep for 1.5 seconds. Minimises chance of getting locked out of PLoS API.
    Sys.sleep(1.5)

    # [I have some code that goes here that processes the search hits (e.g. searching through the Results with a regular expression) in "x" and saves some of the data for later analysis]

}

I chose 300,000 since that is comfortably more than the number of PLOS papers that have been published.

The code kind of works, but it is clear that we are not getting all PLoS papers ever. For example if you change the "for" argument to seq(100000, 300000, num.results), it says " Error: Internal Server Error" and crashes, though one can enter any other numbers than 100,000 and it generally works. Another issue is this:

The x axis shows the dates of papers in PLoS Biology that turned up in my search, expressed as days before today. Looks like there are some ranges that have not been sampled by my current search method. This could be that because of issues with my code that I snipped from the above section (it crashes occasionally when "start" gets over 100,000, and I don't yet know why), but I wanted to check first that our general approach is sensible. Is that what you'd do if you wanted all the Results sections?

If not, is there a simpler way to get the list of fields "f" in my code for all papers on the API? I guess I am not really trying to do a "search", so much as accessing a set of fields in all entries - maybe the function is not built with this task in mind. But I feel that a great addition to rplos would be an easier way of easily getting, say, all the full texts of all PloS papers into R for some text mining.

What other functions should we add to this package?

-Seems like a useful function would be one that graphs results using ID's to label articles so that one can look at the figure, then go get that specific article based on looking at the figure. Probably over my head at the moment. Just plotting results I suppose is a good start.

-What sorts of text mining functions would be interesting?

None of the examples work. e.g.

> plot_throughtime(list("reproducible research"), 500)
Error in `$<-.data.frame`(`*tmp*`, "V1", value = numeric(0)) :
  replacement has 0 rows, data has 500
Calls: plot_throughtime -> $<- -> $<-.data.frame

Jennifer Lin was asking for clarification of parameters, and I noticed that at least toquery isn't accurate as to what it does. Fix up docs, and give examples, etc

From Martin: "We also want to support other identifiers with the "type=doi|pmid|pmcid|mendeley" option, doi is the default, and pmid is particularly interesting."

Right now, only have hl param to do highlighting or not. Add in others. See

http://wiki.apache.org/solr/HighlightingParameters

The same API is also available in a few other places, and we want to use rplos here as well.

Can only pull down up to 999 search results from PLOS. Solution:

Use 'start' argument to start at certain record, then use 'rows' argument which will pull down results of start-to-rows length

should be able to grab the numFound result in the search output from the first set of results, and if the length of search results in less than numFound, we can then pull down the rest of the results in increments of 999 until all are pulled down.

Sound reasonable?

http://api.plos.org/search-examples/schema.xml

e.g, missing fields

• doc_type
• doc_partial_type
• doc_partial_body
• doc_partial_parent_id
• body_rev
• body_ngram
• everything_rev
• everything_ngram
• everything_noprocess
• etc.

Right now the function alm (just changed name from almplosallviews) has four parameters for each of the four ID types, and you can only use of them. It would be nice to be able to detect them automatically, but Pubmed IDs (pmid) and and PubMed Central IDs (pmcid) as far as I can tell can not be disambiguated.

Any thoughts anyone?

https://github.com/hack4ac/plos-author-contributions/blob/master/schema.xml

What is our protocol on documentation pages and where they reside?

Currently we have: 1. The website (a bit out of date in terms of tutorials and examples). 2. Some repos have gh-pages 3. Others have neither.

Should we explore a way to keep this consistent somehow? Update only the github documents and have that mirror on wordpress (plugin?)

thoughts?

at least the examples don't because they don't match the function definition.

Annotation DOIs are now returned along with full paper DOIs when doing fq=doc_type:full, so need to exclude annotation DOIs with fq=-article_type:Correction

E.g., http://api.plos.org/search?q=*:*&fq=doc_type:full&fq=-article_type:Correction&fl=id,title_display,article_type%20desc&api_key=DEMO&wt=json

We should consider which of the use cases we can address from the idea-list from PLoS: http://blogs.plos.org/plos/2011/08/app-ideas-announced-%E2%80%93-develop-one-for-the-binary-battle/

Any idea how to remove those .DS_Store files that are hidden on my local directory, but show up on github repo?

Thanks, S

toquery is misleading, as the actual parameter it represents is fq, which actually filters a query, but doesn't affect the query itself.

Change:

• terms -> q
• fields -> fl
• toquery -> fq

But leave in old params for a good deprecation message

Could be useful, except their API is pretty bad. E.g. no way to do paging, returns 50 results, then you just have to modify your query to get others.

http://www.sherpa.ac.uk/romeo/api.html

There seem to be two issues with the searchplos function:

• result counts are always API result count -1
• the limit of results seems to be limit - 500

CrossRef's API (CrossRef provides the doi functionality that many services use to get the metadata associated with the doi.)
Haven't taken a close look at it yet, but see: http://labs.crossref.org/site/quick_and_dirty_api_guide.html

William Gunn (@MrGunn) on twitter asked "off the top of your head, do you know a good one liner to aggregate counts by day given timestamps including HH:MM:SS?"

@ropensci/owners Hey, curious what your thoughts are on the default output for functions in rplos.

I used to output a data.frame or list of the data, but recently thought it may be better to output a simple S3 class, and have a print method, so what is returned is like

out <- searchplos('ecology', 'id,publication_date', limit = 2)

 Data
No. of records: 2
No. of variables: 2
Variables: id, publication_date

 Highlighting
No. of records: 0
No. of variables: 0
Variables: 

And then they can get data in a data.frame by

plos_todf(out)

                            id     publication_date
1 10.1371/journal.pone.0059813 2013-04-24T00:00:00Z
2 10.1371/journal.pone.0001248 2007-11-28T00:00:00Z

I thought the S3 class seems appropriate since you can quickly get a lot of data back, and its a bit overwhelming perhaps. And I am adding in highlighting and faceting data in the output, so it sorta makes sense I think to do this. And with highlighting and faceting output could do plos_todf(out, return='data'), plos_todf(out, return='highlighting'), or plos_todf(out, return='facets'), or plos_todf(out, return='all').

Thoughts?

When vis=FALSE, need to return only data 🐑

See here http://wiki.apache.org/solr/SimpleFacetParameters

Also, e.g's here http://lucene.apache.org/solr/4_6_0/tutorial.html

William Gunn asked if searches on their APIs could do regular expression searches. I sort of doubt it, but need to get in touch with the PLoS folks to see if it can be done.

Use parsing algorithms from almevents

See martin's functions in plosOpenR almEventFetch and maybe twitterFetch

http://xissn.worldcat.org/xissnadmin/doc/api.htm

see documentation, and email from Joe

Error in data.frame(issn = "0081-685X", journal_title = "Studia Psychologiczne", :
Argumente implizieren unterschiedliche Anzahl Zeilen: 1, 0
Error in data.frame(issn = "2191-0200; 0334-1763", journal_title = "Reviews in the Neurosciences", :
Argumente implizieren unterschiedliche Anzahl Zeilen: 1, 0
Error in x[[1]]$contributors : $ operator is invalid for atomic vectors
Error in data.frame(issn = "1868-1891; 1868-1883", journal_title = "Hormone Molecular Biology and Clinical Investigation", :
Argumente implizieren unterschiedliche Anzahl Zeilen: 1, 0
Error in x[[1]]$contributors : $ operator is invalid for atomic vectors
Error in x[[1]]$contributors : $ operator is invalid for atomic vectors

Add this to the function plot_throughtime, as an additional option for visualizing search results. Output needs to be rendered in a browser, so cannot be visualized locally. Probably need to change data frame format from that used for the ggplot graphics.

At load, get the error

library(rplos)
Loading required package: ggplot2
New to rplos? Tutorial at http://ropensci.org/tutorials/rplos_tutorial.html. Use suppressPackageStartupMessages() to suppress these startup messages in the future
Warning messages:
1: replacing previous import ‘rename’ when loading ‘reshape’
2: replacing previous import ‘round_any’ when loading ‘reshape’

• The options info=event and source=crossref are missing

• Limit for ALM API is 50 articles at a time.

• I also want to add the source=crossref option to almevents.

• The example "almevents(doi="10.1371/journal.pone.0029797", key = 1)" is currently broken,: results in:

  Error in names(temp[[1]]) <- c("bloglines", "citeulike", "connotea", "crossref", :
  Attribut 'names' [18] muss dieselbe Länge haben wie der Vektor [0]

• Enable the year=2012 option.

Add note to documentation and fix in code if needed.

Martin said there is leading and trailing white space in some of the returned data from either the search api or the alm api

And some html in titles and other fields - try to remove those

Hi, I am a new R user.

Forgive me if my question is too naive.

I have tried to install this package, but clearly, the zip file downloaed from the source file is not actually an recognizable R package. Then I unzip the file and set R work directory to it, run the code from wiki, R told me "Error: could not find function "searchplos" "

Could you tell me how to install it in R or how to directly use it?

Thank you very much! ~

Figure out how to suppress the googleVis startup message when loading rplos.

A user can do this suppressPackageStartupMessages(library(googleVis)), but need some way to do it within rplos.

This maybe helpful