Modify retrieval of forecast/advisory products

[timtrice]

As it stands now, the flow of get_fstadv or the equivalent get_storm_data(products = "fstadv") is:

1. Get a character vector (length 1 or more) of links to the storm's archive page to get_storm_data
2. The vector is split into groups of 4 and each group is retrieved. At this point you have all links from each requested storm's archive page.
3. Forecast/Advisory links are extracted and sent to extract_product_contents where the text for each product is extracted.
   • In this example, the text is parsed in fstadv(), loaded into a dataframe and returned.
4. A list of dataframes for each storm is returned.

With this, any cyclone in the AL or EP basin since 1998 has a fstadv product.

These products do exist on the FTP server. However, they are incomplete for the same time period and both basins.

Most recent cyclone fstadv products can be found here: ftp://ftp.nhc.noaa.gov/atcf/mar/

Archived fstadv products can be found here: ftp://ftp.nhc.noaa.gov/atcf/archive/MESSAGES/

However, not all years have these products (and can possibly be assumed not all cyclone's products will exist here, either).

• There is no data at all for 1998, 2000:2002
• Data exists for 1999 but not fstadv products (no "mar" directory)

[timtrice]

More fstadv products may be located in another subdirectory of the FTP server. In the archives (ftp://ftp.nhc.noaa.gov/atcf/archive/) are yearly subdirectories. In 1998, for example, are dat.gz files for each cyclone. It appears,

• a*1998.dat.gz: Computer forecast model data
• b*1998.dat.gz: Best-track data
• f*1998.dat.gz: Seems to be more forecast model data

Then, there are two subdirectories:

• old_fixes: seems to contain fix points by satellite (possible dvorak info)
• messages

It is messages that may contain what I want. For example, there is a al011998_msg.zip. Inside,

• adal0198.num - seems to contain a numeric value for the number of advisories issued for the storm
• lal0198. - where is advisory number; these are strike probabilities
• mal0198.* - forecast/advisory data
• nal0198.* - storm discussion
• pal0198.* - public advisory

So, may be able to use the FTP server to get all data without going to the front-end.

Additionally, it seems the archives for these text products may go all the way back to 1991 giving another 7 years of data

Computer forecast model data may exists for even earlier cyclones.

[timtrice]

Allow either a key or a year (extract the year from key)

Given the year, determine what link to access;

If not current year, ftp://ftp.nhc.noaa.gov/atcf/archive/MESSAGES/2017/

Else, ftp://ftp.nhc.noaa.gov/atcf/

After this point, the functionality should remain the same.

• adv - Contains public advisory data in XML format
• aid_public - appears to be forecast model products
• btk - possibly best track data
• com - possibly more model products
  • tcvitals-arch
  • tcvitals
• dis - storm discussion
• fix - appears to be satellite fixes and comments
• fst - possible forecast positions
• gis
  • 5day
  • btk
  • examples
  • fst
  • kmz
• gpce
• index - info on recent storms
• lsdiag - ?
• mar - forecast advisory text files
• mar-coop - NA
• pns - ?
• pub - public advisories in text files
• stext - SHIPS forecast data
• wndprb - wind probabilities (These may be under a different file where strike probabilities were used).

Example of retrieving FTP list of URLs

url <- "ftp://ftp.nhc.noaa.gov/atcf/archive/MESSAGES/2017/mar/"
hdl <- curl::new_handle(dirlistonly = TRUE)
con <- curl::curl(url, "r", hdl)
tbl <- read.table(con, stringsAsFactors = FALSE, fill = TRUE)
close(con)

[timtrice]

Duplicate more or less of #113