Tidy?

[PMassicotte]

Should we reformat the output as this?

library(gtrendsR)

res <- gtrends("nhl", geo = "CA")

res <- res$trend
head(res)

##        start        end nhl.CA
## 1 2004-01-04 2004-01-10     21
## 2 2004-01-11 2004-01-17     21
## 3 2004-01-18 2004-01-24     22
## 4 2004-01-25 2004-01-31     20
## 5 2004-02-01 2004-02-07     21
## 6 2004-02-08 2004-02-14     21

res2 <- data.frame(start = res$start, 
                   end = res$end, 
                   keyword = "nhl",
                   geo = "CA", 
                   hits = res$nhl.CA)

head(res2)

##        start        end keyword geo hits
## 1 2004-01-04 2004-01-10     nhl  CA   21
## 2 2004-01-11 2004-01-17     nhl  CA   21
## 3 2004-01-18 2004-01-24     nhl  CA   22
## 4 2004-01-25 2004-01-31     nhl  CA   20
## 5 2004-02-01 2004-02-07     nhl  CA   21
## 6 2004-02-08 2004-02-14     nhl  CA   21

[eddelbuettel]

Oh, gosh, I had to squint hard to get at what you wanted here as

R> library(gtrendsR)
Creating a generic function for ‘toJSON’ from package ‘jsonlite’ in package ‘googleVis’
R> res <- gtrends("nhl", geo = "CA")
R> head(res$trend)
       start        end nhl.CA
1 2004-01-04 2004-01-10     21
2 2004-01-11 2004-01-17     21
3 2004-01-18 2004-01-24     22
4 2004-01-25 2004-01-31     20
5 2004-02-01 2004-02-07     21
6 2004-02-08 2004-02-14     21
R> class(res$trend)
[1] "data.frame"

wasn't quite sure what you suggested. In short, "could do" -- but what about more complex queries like the recent 'cloud and big data'. At some point it gets silly to repeat this. A more compact form would be to set an attribute via attr() .... but then all this is already in the res object returned.

So I don't really see yet why we need to stick it into the data.frame. Maybe add as an example, or maybe add a helper function resToTidy() ?

[PMassicotte]

The only thing that "bug" me is nhl.CA where we have both keyword and country in the column name.

BTW, just google made some changes and we can now download CSV by chuck.

https://www.google.com/trends/explore?date=all&q=nhl

[eddelbuettel]

Gotcha. In that case ... shall we just override that with maybe a fixed word count, or hits, or put the (possibly long multi-word) query there?

Download by chunk ... is better for really large result sets?

[PMassicotte]

Alright, I will work on that (the tidy thing).

I will make more sens for request list this:

> head(gtrends("蘋果",geo="TW")$trend)
       start        end X.e8..98..8b..e6..9e..9c..TW
1 2004-01-04 2004-01-10                           20
2 2004-01-11 2004-01-17                           21
3 2004-01-18 2004-01-24                           22
4 2004-01-25 2004-01-31                           17
5 2004-02-01 2004-02-07                           19
6 2004-02-08 2004-02-14                           26

[eddelbuettel]

Absolutely! I like 'hits' as a column number, and a 'long-to-wide' exporter helper may be useful.

But this grumpy old man doesn't fall for every seasonal flavour of the tidyverse so just get off my lawn...

[PMassicotte]

Hehehe do not fear my friend, I did it like this:

  trend <- reshape(
    trend,
    varying = tmp_kw,
    v.names = "hits",
    direction = "long",
    timevar = "keyword",
    times = requested_kw
  )

[eddelbuettel]

No school like the old school. Using stats::reshape is the sign of a true craftsman :)