Results 1 to 4 of 4
  1. #1

    data conversion testing startegy

    I am currently involved with a data-conversion project that involves building a conversion engine (and supporting web application ) that converts XML files to CSV and vise-versa. To test this system we are developing a parallel system (different dev team). Now to verify the validity of this parallel system the project manager is asking me to verify the output/resultant document manually (visual comparison) to check the data conversion rules have been implemented correctly. In order to do the tester would have to open-up the XML the XSD and the CSD (csv schema) and CSV file and verify record by record. The manager says for every class of files one record would have to be verified. This project btw involves conversion amazon.com feeds (XML) to merchant understandable format(CSV). There are about 44 feeds and we would have to verify one record for every feed. All of this is being done to verify the parallel system.

    I on the other hand do not agree with him. I think manual verification is wrong strategy. I think the strategy should be something like this:
    1)we give the same input XML to both systems
    2)compare the output using a text comparison tool
    3)Both dev teams should now arbitrate any conflicts that may have come out in the comparison process.
    4)if no mis-match occurs then move one.

    Basically both developers will put forward the reason why their system is right and finally in this discussion the bug in the system (in the original system or the parallel system ) will be pointed out and resolved.

    BTW the XML structure is quite complex and involves many rules. Please share your view on the strategies.

    Below is the sample XML and XSD. I believe visual comparison is a bad choice and will not give satisfactory/definitive result.

    --------------- XML ------------------------
    <?xml version="1.0" encoding="UTF-8" ?>
    <SmallImage xmlns:aws="http://webservices.amazon.com/AWSECommerceService">
    <aws:Height Units="pixels">50</aws:Height>
    <aws:Width Units="pixels">50</aws:Width>
    ------------- End XML ------------------------------

    Here is my schema file:

    ---------------- Schema -----------------------------
    <?xml version="1.0" encoding="UTF-8"?>
    <xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema"


    <sql:relationship name="Item_Image"
    child-key="vcASIN" />


    <xsd:element name="Item" sql:relation="tblAmazonBase_Load" type="ItemType" />

    <xsd:complexType name="ItemType" >
    <xsd:element name="ASIN" sql:field="vcASIN" type="xsd:string" />
    <xsd:element name="SalesRank" sql:field="intSalesRank" type="xsd:integer" />

    <xsd:element name="SmallImage" sql:relation="tblAmazonImage_Load"
    type="ImageType" />

    <xsd:complexType name="ImageType">
    <xsd:element name="URL" sql:field="vcSmallImageUrl" type="xsd:string" />
    <xsd:element name="Height" sql:field="intSmallImageHeight" type="xsd:integer" />
    <xsd:element name="Width" sql:field="intSmallImageWidth" type="xsd:integer" />

    ------------------- End of Schema ----------------------------------

  2. #2

    Re: data conversion testing startegy

    Personally I do think that some manual validation should be done as well. What if both development teams make the same mistake, or interpret a rule the same way and both systems output the same value but it is wrong... you would catch that if you did a manual validation of X was the input and Y should be the output and look, both systems output Z.

    I do agree that you should do something where you compare the 2 outputs, that is a great way to check on a larger scale.

  3. #3

    Re: data conversion testing startegy

    hi, thanks for your reply.
    I see the following problems in testing the parallel system by manual verification.

    1) One more person would have to understand the mapping and conversion rules.
    2)This new resource has to have the same/similar capability as the team that developed the conversion engine and the team that developed the parallel system. Mapping is not straightforward and could be interpreted in more than one way. This resource I am talking about will be a tester with little or no experience of working with XML.

  4. #4

    Re: data conversion testing startegy

    My suggestion:

    It's counter productive to have an inexperienced person looking for XML issues. Instead, have the developers of both teams review each others code, i.e. conduct a code walkthrough.

    Then do the comparison as you've suggested.




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
BetaSoft Inc.
All times are GMT -8. The time now is 05:45 PM.

Copyright BetaSoft Inc.