summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorThomas Kulik <thomas.kulik@telekom.de>2020-03-19 14:49:59 +0100
committerThomas Kulik <thomas.kulik@telekom.de>2020-03-19 14:51:14 +0100
commit961abe14ea43460cb9fe5b9ec48bb1e276a65d13 (patch)
tree9888e16e0048a330c270fa465aadfd18141102ff
parent613beb9ad4681acbcf4a61eb761eccb185424e18 (diff)
Issue-ID: DOC-594
warnstat update 1.4.0 Signed-off-by: Thomas Kulik <thomas.kulik@telekom.de> Change-Id: I93e8d2bf5a922ec4c4c485a311bf010b48f60ac1
-rwxr-xr-xtools/warnstats.sh156
1 files changed, 136 insertions, 20 deletions
diff --git a/tools/warnstats.sh b/tools/warnstats.sh
index 96bbfb8bc..648aef258 100755
--- a/tools/warnstats.sh
+++ b/tools/warnstats.sh
@@ -25,27 +25,36 @@
### DESCRIPTION:
### warnstat helps to find the onap modules (projects) and rst-files which are
### responsible for the most warnings during the documentation build process
-### it requires a tox build logfile, parses it line by line and prints out some
-### statistics
+### it requires a tox build logfile, parses it line by line, prints out some
+### statistics and provides links to the local rst file and html version.
###
###
### CHANGELOG (LATEST ON TOP)
###
-### 1.3.1 (2020-03-10) fixed minor typo in usage message
-### 1.3.0 (2020-03-09) initial release
+### 1.4.0 (2020-03-18) - the link to the local html and rst file is provided in
+### the output. this may help to ease the debug process.
+### use mouse-over/context menu functionality of bash to
+### easily open files with your browser or rst editor.
+### - improved handling for module names (in case they are
+### no real onap projects/modules but directories which
+### contain additional documentation in rst format).
+### 1.3.1 (2020-03-10) - fixed minor typo in usage message
+### 1.3.0 (2020-03-09) - initial release
###
-script_version="1.3.1 (2020-03-10)"
+script_version="1.4.0 (2020-03-18)"
echo " ";
-echo "warnstats - Version ${script_version}";
-echo " ";
+echo " warnstats version ${script_version}";
declare -A module_array
declare -A message_short_array
declare -A message_long_array
declare -A rstfile_array
+declare -A rstfilepath_array
+declare -A htmlfilepath_array
+declare -A webpath_array
###
### simple script argument handling
@@ -65,23 +74,107 @@ if [ ! -f $logfile ] ; then
exit 1
fi
+# get local html build directory
+html_build_dir=$(grep "sphinx-build -b html" $logfile);
+html_build_dir=$(echo "$html_build_dir" | grep -oP "/ .*/doc/docs/_build/html$");
+html_build_dir=$(echo "$html_build_dir" | sed -r 's:^/ ::');
+echo " tox html build directory: $html_build_dir"
+
# read in the tox build logfile - use only lines which contain a warning
readarray -t logfile_array < <(grep ": WARNING:" $logfile);
# process filtered logfile line by line
+echo " tox logfile: $logfile";
for line in "${logfile_array[@]}"
do
# count warning lines
(( counter++ ));
- echo -n -e "lines processed: $counter\r";
+ echo -n -e " lines processed: $counter\r";
+
+ #
+ # extract path to local rst file
+ #
+ path_rst=$line;
+ #echo "DBUG line: $line"
+ # remove problematic text in line that causes regex to fail
+ path_rst=$(echo "$path_rst" | sed -r 's:, other instance in.*::');
+ #echo "DBUG path_rst: $path_rst"
+ # grep the rst file path
+ path_rst=$(echo "$path_rst" | grep -oP "^/.*\.rst");
+ #echo "DBUG path_rst: $path_rst"
+ if [[ "$path_rst" == "" ]] ; then
+ path_rst="path_to_rst_missing"
+ #echo "DBUG path_rst: $path_rst"
+ fi
+ path_rst="file:$path_rst";
+ #echo "DBUG path_rst: $path_rst"
+ # finally embed the full rst path in a message to use mouse-over/context menu of bash to open file
+ #echo -e '\e]8;;'$path_rst'\a(rst)\e]8;;\a'
+ path_rst='\e]8;;'$path_rst'\arst\e]8;;\a';
+ #echo -e "DBUG path_rst: "$path_rst;
+
+ #
+ # extract path to the html version of the local rst file
+ #
+ path_html=$line;
+ #echo "DBUG line: $line"
+ # remove problematic text in line that causes regex to fail
+ path_html=$(echo "$path_html" | sed -r 's:, other instance in.*::');
+ #echo "DBUG path_html: $path_html"
+ # grep the rst file path and modify it so we get the local html build path; grep a little bit more to be save
+ path_html=$(echo "$path_html" | grep -oP "(^|/)docs(/.*|)/[\w -]*\.rst");
+ #echo "DBUG path_html: $path_html"
+ path_html=$(echo "$path_html" | sed -r 's:^/docs::');
+ #echo "DBUG path_html: $path_html"
+ path_html=$(echo "$path_html" | sed -r 's:.rst:.html:');
+ #echo "DBUG path_html: $path_html"
+ # create also the path to the web version
+ path_web="https://docs.onap.org/en/latest$path_html"
+ path_web='\e]8;;'$path_web'\aweb\e]8;;\a';
+ #echo "DBUG path_web: $path_web"
+ path_html="file:$html_build_dir$path_html";
+ #echo "DBUG path_html: $path_html"
+ # finally embed the full html path in a message to use mouse-over/context menu of bash to open file
+ #echo -e '\e]8;;'$path_html'\a(html)\e]8;;\a'
+ path_html='\e]8;;'$path_html'\ahtml\e]8;;\a';
+ #echo -e "DBUG path_html: "$path_html;
+
+
+ # extract module name from line (remove all text before module name; then cut out module name)
+ module=$(echo "$line" | sed -r 's:(^.*/doc/docs/submodules/|^docs/submodules/|checking consistency... )::' | cut -f1 -d\/);
+ #echo "DBUG line: $line"
+ #echo "DBUG module: $module"
+
+ # in case the extraction has not lead to a valid module name do some additional investigation
+ if [[ "$module" == "" ]] ; then
- # extract module name from line
- module=$(echo "$line" | sed -r 's:^/.+/doc/docs/(submodules|guides)/::' | cut -f1 -d\/);
+ if [[ $line =~ doc/docs/release ]] ; then
+ module="<docs/release>"
+ #echo "DBUG line: $line"
+ #echo "DBUG module: $module"
+ elif [[ $line =~ doc/docs/use-cases ]] ; then
+ module="<docs/use-cases>"
+ #echo "DBUG line: $line"
+ #echo "DBUG module: $module"
+ elif [[ $line =~ doc/docs/guides ]] ; then
+ module="<docs/guides>"
+ #echo "DBUG line: $line"
+ #echo "DBUG module: $module"
+ else
+ module="<docs>"
+ #echo "DBUG line: $line"
+ #echo "DBUG module: $module"
+ fi
- # in case the extraction has no valid name fill the missing field
- if [[ "$module" == "" ]] ; then
- module="<missing_module_name>";
fi
+ #echo "DBUG line: $line";
+ #echo "DBUG module: $module";
+
+ # get the maximum length of the variable entries to adjust table width later on
+ if [[ ${#module} -gt "$maxlength_module" ]]; then
+ maxlength_module=${#module};
+ fi
+ #echo "DBUG maxlength_module=$maxlength_module";
# extract rst file name from line and do some formatting to use it later as an array name
#echo "DBUG line: $line";
@@ -89,13 +182,24 @@ do
rstfile=$(echo -e ${rstfile} | tr '[:blank:]' '_');
#echo "DBUG rst-file: $rstfile";
+ # get the maximum length of the variable entries to adjust table width later on
+ if [[ ${#rstfile} -gt "$maxlength_rstfile" ]]; then
+ maxlength_rstfile=${#rstfile};
+ fi
+ #echo "DBUG maxlength_rstfile=$maxlength_rstfile";
+
# count the number of warnings for the module/rstfile combination
- (( rstfile_array[$module | $rstfile]++ ));
+ (( rstfile_array[$module | $rstfile]++ ));
# count the number of warnings for the single module
- #echo "DBUG $module | $rstfile | $message";
+ #echo "DBUG $module | $rstfile | $message";
(( module_array[$module]++ ));
+ # now we have all the information to fill the html/rst/web (file) path arrays
+ htmlfilepath_array[$module | $rstfile]=$path_html;
+ rstfilepath_array[$module | $rstfile]=$path_rst;
+ webpath_array[$module | $rstfile]=$path_web;
+
# extract the warning message and do some formatting
#message=$(echo "$line" | sed -r 's:^/.+WARNING\:\ ::');
message=$(echo "$line" | sed -r 's:^.+WARNING\:\ ::');
@@ -137,7 +241,7 @@ do
((nc += n))
#format counter to have always x digits
n=$(printf "%05d" $n);
- echo " $n | $m" >>tempoutfile;
+ echo " $n | $m" >>tempoutfile;
done
#format counter to have always x digits
@@ -165,7 +269,7 @@ do
((nc += n))
#format counter to have always x digits
n=$(printf "%05d" $n);
- echo " $n | $m" >>tempoutfile;
+ echo " $n | $m" >>tempoutfile;
done
#format counter to have always x digits
@@ -192,7 +296,7 @@ do
n=${module_array[$i]};
((nc += n))
n=$(printf "%05d" $n);
- echo " $n | $m" >>tempoutfile;
+ echo " $n | $m" >>tempoutfile;
done
#format counter to have always x digits
@@ -215,10 +319,19 @@ for i in "${!rstfile_array[@]}"
do
m=$i;
n=${rstfile_array[$i]};
+ p=${htmlfilepath_array[$i]}
+ r=${rstfilepath_array[$i]}
+ w=${webpath_array[$i]}
+ #echo "DBUG -------------------------------"
+ #echo "DBUG i=$i"
+ #echo "DBUG m=$m"
+ #echo "DBUG n=$n"
+ #echo "DBUG p=$p"
+ #echo -e "DBUG p=$p"
((nc += n))
#format counter to have always x digits
n=$(printf "%05d" $n);
- echo " $m | $n" >>tempoutfile;
+ echo -e " $m | ($r,$p,$w) | $n" >>tempoutfile;
done
#format counter to have always x digits
@@ -244,10 +357,13 @@ for i in "${!rstfile_array[@]}"
do
m=$i;
n=${rstfile_array[$i]};
+ p=${htmlfilepath_array[$i]}
+ r=${rstfilepath_array[$i]}
+ w=${webpath_array[$i]}
((nc += n))
#format counter to have always x digits
n=$(printf "%05d" $n);
- echo " $n | $m" >>tempoutfile;
+ echo -e " $n | $m | ($r,$p,$w)" >>tempoutfile;
done
#format counter to have always x digits