Problem Parsing CSV File in GEL Script

[Alan Brobst]

I'm a relative newcomer to GEL and not a Java guru, so I'm hoping someone out there can give me some advice.

 

My goal is to use GEL to read a CSV file (which is exported from our ERP system and stored on our SFTP server), then parse each row into individual columns, and then populate a staging table via XOG.

 

I have almost everything working exactly right except for one problem: the content of the CSV file is giving me headaches.  It's comma-delimited, but the only time quotation marks are used is when there is a comma in a field that is not a delimiter.  In those cases, the value is embedded in quotation marks.  For example:

 

PART_NUMBER,ITEM_DESCRIPTION,UNIT_COST,QTY_SHIPPED
3EU48C891,CIRCUIT BREAKER BOX,138.22,50
4FJ62D837,SWITCH PLATE,0.04,5000
5GL29E906,"BREAKER, CIRCUIT, 20A",6.94,200

 

Note that in the first 2 rows the value in the second column DOES NOT contain commas, so there are no quotation marks around the ITEM_DESCRIPTION field. In the third row, the value in the second column DOES contain commas, so there are quotation marks before and after the ITEM_DESCRIPTION value itself.

 

When I parse the row using files:readFile coded as follows:

 

<files:readFile fileName="${vFileName}" delimiter="," var="vPartData" embedded="false"/>

 

I get:

 

First Row: (OK)
Field 1 = 3EU48C891
Field 2 = CIRCUIT BREAKER BOX
Field 3 = 138.22
Field 4 = 50

 

Second Row: (OK)
Field 1 = 4FJ62D837
Field 2 = SWITCH PLATE
Field 3 = 0.04
Field 4 = 5000

 

Third Row: (NOT OK)
Field 1 = 5GL29E906
Field 2 = "BREAKER
Field 3 = CIRCUIT
Field 4 = 20A"

 

The fields:readFile doesn't recognize that the commas in the second field in the third row aren't delimiters, and it parses the third row incorrectly.  The ITEM_DESCRIPTION field becomes 3 separate columns because my command doesn't recognize that the 2 commas in that value aren't supposed to be delimiters.

 

Does anyone know the proper delimiter syntax so that all rows will be parsed correctly, even if there are non-delimiting commas in a row?  If you do, you will be my hero.

 

Thanks,

 

Alan Brobst

GE Transportation, a Wabtec Company

[Roland Parrotte]

I generally use | as the column delimiter, when reading in delimiter files.  Yes, the comma delimited file should work (and someone else may be able to provide a solution).  The | character is generally not used in text fields, and hence I don't normally hit the issue you are currently experiencing.

[David Morton]

Have you tried embedded="true" - thats the flag about the fields being enclosed in quotes ; but I'm not sure if that would break when (normally) the field is not enclosed in quotes.

 

(just GUESSing)

[Alan Brobst]

Hi David,

 

Thanks for that thought, but unfortunately, it didn't work.  When I changed embedded="false" to embedded="true" in my files.readFile line, it wouldn't return any output.

 

Alan

[Eugene Greiff]

I suspect the embed attribute is going to want each field wrapped in double quotes but I would give Dave’s idea a try.

 

In looking at your data, it appears that only the second field is of concern.

 

Another option would be to check the length of the row items (I believe it is a string[] type) and if equal to 4 then life is good.  If greater than four you would need to take the first and last two elements of the item array and then combine the elements from 1 to length -3 (zero based).

 

V/r,

Gene

[Alan Brobst]

Hi Gene,

 

Thanks for that suggestion.  My bigger issue is that the number of commas in the second field is inconsistent.  Some values might have 1 comma, some might have 5, etc.  And there are 2 others fields in the source CSV file that can also contain commas, so the permutations are daunting.

 

Thanks,

 

Alan

[Roland Parrotte]

Left field suggestion, have you considered generating the data as XML, and reading the XML file via the GEL script?

[Eugene Greiff]

How about manual parsing of the CSV file.

 

Multiple fields that contain ",,," data elements.

 

<?xml version="1.0" encoding="utf-8"?>
<gel:script
     xmlns:core="jelly:core"
     xmlns:log="jelly:log"
     xmlns:util="jelly:util"
     xmlns:files="jelly:com.niku.union.gel.FileTagLibrary"
     xmlns:gel="jelly:com.niku.union.gel.GELTagLibrary"
     xmlns:xsd="http://www.w3.org/2001/XMLSchema"
     xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">

     <gel:log>Start this Script</gel:log>

     <core:set var="dataFile" value="gel\TestData.txt"/>
     <gel:log>dataFile ${dataFile}</gel:log>
     <util:file var="fileOne" name="${dataFile}" />
     <util:available file="${fileOne}">
          <gel:log>${dataFile} file is available</gel:log>
     </util:available>

     <files:readFile fileName="${dataFile}" delimiter='"' var="lines" embedded="false"/>
     <core:forEach var="line" items="${lines.rows}" indexVar="rowCnt">
          <core:if test="${rowCnt != 0}">
               <gel:log>Processing ${rowCnt}</gel:log>
               <util:tokenize var="tokens" delim=""/>
               <core:forEach var="field" items="${line}">
                    <core:choose>
                         <core:when test='${field.startsWith(",") || field.endsWith(",")}'>
                              <core:set var="items" value='${field.split(",")}' />
                              <core:forEach var="thisField" items="${items}">
                                   <core:if test="${!thisField.isEmpty()}" >
                                        <core:expr value="${tokens.add(thisField)}" />
                                   </core:if>
                              </core:forEach>
                         </core:when>
                         <core:otherwise>
                              <core:expr value="${tokens.add(field)}" />
                         </core:otherwise>
                    </core:choose>
               </core:forEach>
               <gel:log></gel:log>
               <core:forEach begin="0" end="${tokens.size()-1}" step="1" indexVar="index">
                    <gel:log>arraylist tokens = ${tokens.get(index)}</gel:log>
               </core:forEach>
          </core:if>
     </core:forEach>
     <gel:log>End this Script</gel:log>

</gel:script>

 

Results:

 

 

V/r,

Gene

[Alan Brobst]

Thanks, Gene!  You are my hero.

[Alan Brobst]

Hi Gene,

You're still my hero, but I encountered an issue -- if the input file contains null values as represented by consecutive commas (e.g., in the CSV content there's something like this):

1,,CIRCUIT BREAKER,4,6.29

Instead of parsing tokens as [0] = 1, [1] = null, [2] = CIRCUIT BREAKER, [3] = 4 and [4] = 6.29

It parses:

[0] = 1, [1] = CIRCUIT BREAKER, [2] = 4, [3] = 6.29, [4] = null

That is, they're "shifted 1 to the left" where the null occurs.

One option would be to substitute ,null, for every occurrence of ,, (e.g., something like a str.replaceAll()) before it splits the line into tokens.  But I'm not smart enough about Java or GEL to figure that one out.  And I'm sure you have a better idea.

Any suggestions you can provide would be much appreciated!

Thanks,

Alan