how to massage or format html in order to parse with xmstarlet?

Question

Do I first need to run html in the wild through something like jsoup? Not to make it valid in a human sense, might turn it to gibberish, but at least so that xmlstarlet can process the file?

Preferably looking for a CLI which can be installed and used like so:

massage foo.html > bar.xhtml

or at least something along those lines.

Use case:

thufir@doge:~/.html$ 
thufir@doge:~/.html$ curl http://int.soccerway.com/  > soccer.html
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  188k    0  188k    0     0   313k      0 --:--:-- --:--:-- --:--:--  313k
thufir@doge:~/.html$ 
thufir@doge:~/.html$ xmlstarlet sel -t -v "/html/body/table/tr/td[1]" -n soccer.html 
soccer.html:70.13: xmlParseEntityRef: no name
if (this.$ && this.$.fn && this.$.fn.jquery) {
            ^
soccer.html:70.14: xmlParseEntityRef: no name
if (this.$ && this.$.fn && this.$.fn.jquery) {
             ^
soccer.html:70.26: xmlParseEntityRef: no name
if (this.$ && this.$.fn && this.$.fn.jquery) {
                         ^
soccer.html:70.27: xmlParseEntityRef: no name
if (this.$ && this.$.fn && this.$.fn.jquery) {
                          ^
soccer.html:198.8: Opening and ending tag mismatch: link line 27 and head
</head>
       ^
soccer.html:209.45: EntityRef: expecting ';'
  j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src=
                                            ^
soccer.html:223.40: xmlParseEntityRef: no name
      if (typeof(e.data) === 'string' && (e.data.indexOf('onEplayerVideoStarted'
                                       ^

Ideally would run htmlstarlet directly against the URL but there seems no such provision.

there is a fo option to format, but I wasn't able to get different results than above.

Answer 1

If you just want the table's data cells, it is possible using xmlstarlet fo followed by xmlstarlet sel. The main issue you are having is with the XPath. If you add a couple of "wildcard" elements (//), you get the desired result:

# fetch URL silently, following redirects, and send to standard out
curl -sL http://int.soccerway.com/                          |

# interpret input as HTML (-H) and try to recover as much as possible (-R)
xmlstarlet fo  -H -R                           2> /dev/null |

# use the following XPath expression and return the value (-t -v), 
# also add a newline after the result (-n)
xmlstarlet sel -t -v '//table//tr//h3/span' -n 2> /dev/null |

# only show the first 10 values
head -n10

Output:

World - Friendlies
Argentina - Prim B Nacional
Australia - National Premier Leagues
Australia - NPL Youth League
Bangladesh - Premier League
Belarus - Premier League
Benin - Championnat National
Brazil - Serie A
Brazil - Serie D
Brazil - Copa Paulista