Am 02.03.2021 um 17:04 schrieb Sven Schreiber: > # now the real thing > web = > readfile("https://opendata.dwd.de/climate_environment/CDC/observations_germany/climate/monthly/kl/historical/") > > print web  # very long string! (HTML source) > eval xmlget(web, "//a")    # error: xmlParseMemory returned NULL > > </hansl> > > Does the error mean that the page's code is simply too much / too long? > OK, turns out the main problem doesn't have anything to do with gretl's xmlget function. The web page above uses the (valid) HTML 4 element <hr> which isn't valid xml and therefore libxml chokes on this (and by implication then also xmlget). However, there's still something strange with the printout. If I run the following short script: <hansl> clear string web = readfile("https://opendata.dwd.de/climate_environment/CDC/observations_germany/climate/monthly/kl/historical/") web = strsub(web, "<hr>", "") # print web eval xmlget(web, "//a") </hansl> then there's no error, but the "eval" line only produces 40 lines or so, with the last printed line being truncated like this ("_hist.zip" is missing": monatswerte_KL_00183_19360101_20191231 If instead I uncomment the "print web" line above then I get everything, which is several hundreds of lines. Not sure why I need to print the string first! This is with a recent snapshot. But note: If I don't do "eval" but assign the xmlget result to a string variable, everything seems fine, so no deep problem here.