execution.log

$ ./to-csv.sed records.xml 
111111H2,111AA2026
111111N1,111AA2026
111111Q1,111AA2026
111111U1,111AA2026
111111Z1,111AA2026

records.xml

<record name="111111H2" />
<items>
  <field name="Electronic Identifier" value="1"/>
  <field name="Symbol" value="111AA2026"/>
  <field name="Full Symbol" value="111AA202622MARFUT"/>
  <field name="System Identifier" value="1"/>
  <field name="System Identifier Description" value="Description"/>
</items>
<record name="111111N1" />
<items>
  <field name="Electronic Identifier" value="2"/>
  <field name="Symbol" value="111AA2026"/>
  <field name="Full Symbol" value="111AA202621JULFUT"/>
  <field name="System Identifier" value="2"/>
  <field name="System Identifier Description" value="Description"/>
</items>
<record name="111111Q1" />
<items>
  <field name="Electronic Identifier" value="3"/>
  <field name="Symbol" value="111AA2026"/>
  <field name="Full Symbol" value="111AA202621AUGFUT"/>
  <field name="System Identifier" value="3"/>
  <field name="System Identifier Description" value="Description"/>
</items>
<record name="111111U1" />
<items>
  <field name="Electronic Identifier" value="4"/>
  <field name="Symbol" value="111AA2026"/>
  <field name="Full Symbol" value="111AA202621SEPFUT"/>
  <field name="System Identifier" value="4"/>
  <field name="System Identifier Description" value="Description"/>
</items>
<record name="111111Z1" />
<items>
  <field name="Electronic Identifier" value="5"/>
  <field name="Symbol" value="111AA2026"/>
  <field name="Full Symbol" value="111AA202621DECFUT"/>
  <field name="System Identifier" value="5"/>
  <field name="System Identifier Description" value="Description"/>
</items>

to-csv.sed

#!/usr/bin/env sed -nEf

# -n flag to disable printing of unhandled lines
# -E flag to enable extended regular expressions
# -f flag to tell sed that the contents of this file are commands to execute

#
# sed uses two internal "variables" called the pattern space and the hold
# space. Each line read from input is stored inside the pattern space and
# all [s]ubstitution commands are performed on the pattern space. So this
# pattern space is effectively rewritten whenever sed reads a new line of
# input. If we want to remember things between read lines, we need to use
# the hold space. There are a few commands that allows us to move things
# between the two spaces, but in what follows, think of the two spaces as
# two variables with different lifetimes:
#
#   1. pattern space: remembers things during the processing of a line
#   2. hold space: remembers things even as we move across lines
#

# For all lines matching "<record" do all of the bracket-enclosed actions.
/<record/ {
  # The matched line is now store inside the pattern space.

  # Now, replace the read line with just the value of the name attribute,
  # so the pattern space contains just that attribute value. Makes use of
  # the \1 capturing group reference.
  s/<record name="([^"]+)" \/>/\1/

  # Set the value of the hold space to the value of the pattern space, i.e.,
  # the attribute value, because we need to read a few lines before doing
  # additional processing on it.
  h

  # Read the next three lines after the matched line. Each "n" command will
  # set the contents of the pattern space to the contents of the read line.
  n
  n
  n

  # The pattern space now contains the 3rd line after our initial "<record".

  # Keep only the attribute value: \1 is again a capturing group reference.
  s/^ *<field name="[^"]+" value="([^"]+)"\/>/\1/

  # Append a newline and then the pattern space to the hold space. This ensure
  # the order of the string in the hold space is: <record>\n<field>
  H

  # The hold space now contains the name of the "record" tag, a newline and
  # the name of the 3rd "field" element, so we can copy it back from there
  # to the pattern space in order to perform a final substitution and print
  # the result.
  g

  # Replace newlines with commas and [p]rint the result after substitution.
  s/\n/,/p
}