在单引号内转义双引号

Question

for x in $(cat raw_tables.txt)
do
echo '{
    "type" : "jdbc",
    "jdbc" : {
        "url" : "jdbc:mysql://localhost:3306/test",
        "user" : "root",
        "password" : "<pass>",
        "sql" : "select * from "'$x'"",
        "elasticsearch" {
                "cluster" : "Search",
                "host": "<ip>",
                "port": 9300
        },
        "index" : ""'$x'"",
        "type": ""'$x'"" 
    }
}' | java \
   -cp "/etc/elasticsearch/elasticsearch-jdbc-2.3.3.1/lib/*" \
   -Dlog4j.configurationFile=/etc/elasticsearch/elasticsearch-jdbc-2.3.3.1/bin/log4j2.xml \
   org.xbib.tools.Runner \
   org.xbib.tools.JDBCImporter


cat raw_tables.txt
table1
table2
table3

When i run that it comes out as

"index" : ""$x"",

I need it to come out as "index" : "$x", I can't bypass the double quote it's producing and if I try escaping the entire thing with a single quote the script thinks it's stopped.

I've tried everything.. Anything would be appreciated

Thank you!

Answer 1

Use a here-document, this way you don't have to care about quoting (in this case, since the document doesn't need to contain any backticks or dollar signs):

while read x; do
  java ... lots of options \
    more options for java \
    and more options for java  <<END_DOC
{
    "type" : "jdbc",
    "jdbc" : {
        "url" : "jdbc:mysql://localhost:3306/test",
        "user" : "root",
        "password" : "<pass>",
        "sql" : "select * from $x",
        "elasticsearch" {
                "cluster" : "Search",
                "host": "<ip>",
                "port": 9300
        },
        "index" : "$x",
        "type": "$x" 
    }
}
END_DOC
done <raw_tables.txt

The end marker ( END_DOC in this case) needs to be flush to the left, with no indentation.

Answer 2

You shouldn't try to generate dynamic JSON like this, because you don't know if anything in the contents of x needs to be escaped (although in this case, a valid SQL table name is unlikely to need any special treatement). You should use something like jq to generate the JSON instead.

# This is not JSON itself; it is a filter to be used by jq
# to *generate* JSON.
template='{
        "type" : "jdbc",
        "jdbc" : {
          "url" : "jdbc:mysql://localhost:3306/test",
          "user" : "root",
          "password" : "<pass>",
          "sql" : "select * from \($table)",
          "elasticsearch": {
            "cluster" : "Search",
            "host": "<ip>",
            "port": 9300
          },
          "index" : $table,
          "type": $table
        }
      }'

while IFS= read -r x; do
  jq -n --arg table "$x" "$template" | java ...
done < raw_tables.txt 

Using jq solves many of the same quoting problems that a here document does as well.

---

You can also read the table names from standard input, instead of passing them as a JSON variable. You replace $table with . in the template, then call jq with the -R option instead of the --arg option.

template='{
    "type" : "jdbc",
    "jdbc" : {
      "url" : "jdbc:mysql://localhost:3306/test",
      "user" : "root",
      "password" : "<pass>",
      "sql" : "select * from \(.)",
      "elasticsearch": {
        "cluster" : "Search",
        "host": "<ip>",
        "port": 9300
      },
      "index" : .,
      "type": .
    }
  }'

while IFS= read -r x; do
    echo "$x" | jq -R "$template" | java ...
done < raw_tables.txt

Answer 3

here you go, remove one of the double quotes in each pair around $x

$ for x in table1 table2 table2; do
> echo 'bla bla bla
>       "index" : "'$x'"
>       bla bla bla'
> done
bla bla bla
      "index" : "table1"
      bla bla bla
bla bla bla
      "index" : "table2"
      bla bla bla
bla bla bla
      "index" : "table2"
      bla bla bla