From Newsgroup: comp.lang.cobol
In March 2017, there was a discussion relating to decision
tables and EVALUATE. It got me thinking. I searched for
references and located a document at Object Management
Group (OMG) titled Decision Model and Notation (DMN).
<
http://www.omg.org/spec/DMN/1.1/PDF > [a PS version is
available]. The PDF is 182 pages. Section 8 discusses
decision tables.
This document makes clear there are two orientations for
decision tables: vertical and horizontal. Verticle is what
is, perhaps, more common and understood. In fact, many of
the results for a search will return information on
verticle decision tables, as if it was the only orientation.
Since the horizontal orientation is closer to EVALUATE and
is, in most respects, similar to a spreadsheet; I decided
to see whether a spreadsheet could be used to create a
decision table that could be translated into an EVALUATE
statement. No surprise, it can!
I exported the spreadsheet to a .csv file and wrote a
COBOL program <g> to handle the translation. Unfortunately,
.csv files have some formatting issues to deal with. I
decided to use a decision table <g> to deal with those.
This is the final .csv that was input into the program
that uses the library text. Needless to say, 'bootstrapping'
was involved!
-----
CSV-DT,,,,,,,,,
,state,m = fld-start,csv-rec (m:1),csv-rec (m + 1:1),,,,, 1,1,TRUE,quote,any,move 2 to state,add 1 to m,,,
2,1,TRUE,space,"""'""",add 1 to m,,,,
3,1,any,"not "",""",any,move csv-rec (m:1) to ws-rec (n:1),add 1 to m,add 1 to n,,
4,1,any,""",""",any,move fld-sep to ws-rec (n:1),add 1 to m,add 1 to n,move m to fld-start,
5,2,any,not quote,any,move csv-rec (m:1) to ws-rec (n:1),add 1 to m,add 1 to n,,
6,2,any,quote,quote,move quote to ws-rec (n:1),add 2 to m,add 1 to n,, 7,2,any,quote,""",""",move fld-sep to ws-rec (n:1),add 2 to m,add 1 to n,move m
to fld-start,move 1 to state
8,2,any,quote,""" """,add 2 to m,add 1 to n,move m to fld-start,move 1 to state,
-----
The .csv was created using OpenOffice.org CALC so it
should load into LibreOffice and did load into Google
Sheets. Copy, Paste, Save, and import to see how it
looks in a spreadsheet.
The final library text (copybook) is:
-----
*> CSV-DT
evaluate state also m = fld-start also csv-rec (m:1)
also csv-rec (m + 1:1)
*> Rule 1
when 1 also TRUE also quote also any
move 2 to state
add 1 to m
*> Rule 2
when 1 also TRUE also space also "'"
add 1 to m
*> Rule 3
when 1 also any also not "," also any
move csv-rec (m:1) to ws-rec (n:1)
add 1 to m
add 1 to n
*> Rule 4
when 1 also any also "," also any
move fld-sep to ws-rec (n:1)
add 1 to m
add 1 to n
move m to fld-start
*> Rule 5
when 2 also any also not quote also any
move csv-rec (m:1) to ws-rec (n:1)
add 1 to m
add 1 to n
*> Rule 6
when 2 also any also quote also quote
move quote to ws-rec (n:1)
add 2 to m
add 1 to n
*> Rule 7
when 2 also any also quote also ","
move fld-sep to ws-rec (n:1)
add 2 to m
add 1 to n
move m to fld-start
move 1 to state
*> Rule 8
when 2 also any also quote also " "
add 2 to m
add 1 to n
move m to fld-start
move 1 to state
end-evaluate
.
-----
Rules:
1. Detects a quoted field.
2. Removes an extraneous space, since, apparently, an
apostrophe cannot start a field in a speadsheet.
3. Copies one character in an unquoted field.
4. Detects an end of an unquoted field. (Last field
detection is not required, since, at most, one space
character will be added.)
5. Copies one character in a quoted field.
6. Converts two quotation marks to one.
7. Detects an end of a quoted field.
8. Detects an end of a quoted field when it is the last
field.
Given the ability of current spreadsheets to prevent rows
and columns from scrolling, it should IMHO be possible to
create very large decision tables that can be translated
into EVALUATE statements.
SEEN-BY: 154/30 2320/100 0 1 227/0