If you are having a problem with Daffodil, and think that perhaps you have found a bug, then we suggest you:
- check JIRA to see if the bug is already reported
- create a TDML file for reproducing the bug/issue
- create a JIRA ticket and attach your TDML
- email the issue to email@example.com (you have to sign up for that list to be able to send to it.)
- ask in the Daffodil XMPP Chat Room
Some users are working on DFDL schemas for company-confidential, or military For-Official-Use-Only data formats (such as NATO STANAG 5516). For these users, we have a non-public support mailing address firstname.lastname@example.org which goes to the support team at Tresys.
Check JIRA to See if your Issue is Already There
First you should give a search of our JIRA tickets to see if the problem is already recorded.
Here's a list of all tickets about bugs, new features, and improvements, in reverse chronological order (most recent first). You may want to change the issue type, or status specifications to narrow down the list, but most commonly you would just put some search keywords into the search box.
If you do find a bug or a closely related issue that is open status, then you can add your information to it as a comment if you prefer, rather than creating a new issue. Just knowing that another person has run into the issue is helpful at assigning fix priorities.
If you do not easily find an issue, either create a new JIRA bug, email the issue to email@example.com, or ask about it in the Daffodil XMPP Chat Room. Additionally, creating a TDML file can greatly help us reproduce the issue and resolve it.
Create a TDML File that Illustrates the Issue
A TDML file is often useful just to ask a question about how something in DFDL works, for example, to get a clarification. It allows for a level of precision that is often lacking, but also often required when discussing complex data format issues.
The absolutely best way to report a bug is by creating a TDML test file that demonstrates the problem.
TDML stands for "Test Data Markup Language". It is a way of specifying a DFDL schema, the test data, the expected result or expected error/diagnostic messages, and it is all in a single self-contained XML file. IBM started TDML to capture tests for their own DFDL implementation. Daffodil latched onto this and has since extended it a bit, though there is an effort to reconcile TDML dialects so that all implementations can run the same tests.
By convention, a TDML file uses file extension ".tdml" or, more commonly now ".tdml.xml" which enables the TDML "tutorial" features to work.
Below is an annotated TDML file for a very simple example:
<?xml version="1.0" encoding="ASCII"?> <tdml:testSuite suiteName="Bug Report TDML Template" description="Illustration of TDML for bug reporting." xmlns:tdml="http://www.ibm.com/xmlns/dfdl/testData" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xml="http://www.w3.org/XML/1998/namespace" xmlns:dfdl="http://www.ogf.org/dfdl/dfdl-1.0/" xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:ex="http://example.com" xmlns:gpf="http://www.ibm.com/dfdl/GeneralPurposeFormat" xmlns:daf="urn:ogf:dfdl:2013:imp:opensource.ncsa.illinois.edu:2012:ext" xmlns="http://www.w3.org/1999/xhtml" xsi:schemaLocation="http://www.ibm.com/xmlns/dfdl/testData tdml.xsd" defaultRoundTrip="false"> <!-- This example TDML file is for a self-contained bug report. It shows a parse test, and similar unparse test. --> <tdml:defineSchema name="bug01Schema" elementFormDefault="unqualified"> <xs:import namespace="http://www.ibm.com/dfdl/GeneralPurposeFormat" schemaLocation="IBMdefined/GeneralPurposeFormat.xsd" /> <dfdl:defineFormat name="myFormat"> <dfdl:format ref="gpf:GeneralPurposeFormat" lengthKind="implicit" representation="text" encoding="ASCII" initiator="" terminator="" separator="" /> </dfdl:defineFormat> <dfdl:format ref="ex:myFormat" /> <xs:element name="myTestRoot" type="xs:dateTime" dfdl:calendarPattern="MM.dd.yyyy 'at' HH:mm:ssZZZZZ" dfdl:calendarPatternKind="explicit" dfdl:lengthKind="delimited" dfdl:terminator="%NL;" /> </tdml:defineSchema> <tdml:parserTestCase name="dateTimeTest" root="myTestRoot" model="bug01Schema" description="A hypothetical bug illustration about parsing a date time."> <tdml:document> <tdml:documentPart type="text" replaceDFDLEntities="true"><![CDATA[04.02.2013 at 14:00:56 GMT-05:00%LF;]]></tdml:documentPart> </tdml:document> <tdml:infoset> <tdml:dfdlInfoset> <ex:myTestRoot>2013-04-02T14:00:56.000000-05:00</ex:myTestRoot> </tdml:dfdlInfoset> </tdml:infoset> </tdml:parserTestCase> <tdml:unparserTestCase name="unparseDateTimeTest" root="myTestRoot" model="bug01Schema" description="Another bug illustration, this time unparsing direction."> <tdml:infoset> <tdml:dfdlInfoset> <ex:myTestRoot>2013-04-02T14:00:56.000000-05:00</ex:myTestRoot> </tdml:dfdlInfoset> </tdml:infoset> <tdml:document> <tdml:documentPart type="text" replaceDFDLEntities="true"><![CDATA[04.02.2013 at 14:00:56-05:00%CR;%LF;]]></tdml:documentPart> </tdml:document> </tdml:unparserTestCase> </tdml:testSuite>
Suppose you save the above out as a file "myDateTimeBug.tdml". You can then run it using the Daffodil Command Line Interface.
daffodil test myDateTimeBug.tdml
The Infoset element that contains the expected result may need to contain characters that are not legal in XML documents. Daffodil remaps these characters into legal XML characters in the Unicode Private-use-area. See Daffodil and the DFDL Infoset for details.
Specifying Data in Text, Hex, Bits, or External File
When specifying the test data, there are other ways to do this than using just text.
You can specify the test data in hexadecimal, in individual bits, or you can direct Daffodil to find the data in an external file.
These are illustrated here. You just change the way the tdml:document element is specified to include tdml:documentPart children elements:
<tdml:document> <!-- A document part with type="text" is text. Use CDATA to avoid whitespace changes. So in the example below, the line ending after '250;' and after '967;' are intentional parts of the data so as to illustrate that the whitespace is preserved if immportant when you use CDATA bracketing. If you care exactly which kind of line ending is used, then you can use DFDL character entities to insert a %CR; %LF; or both. In this example, because the whitespace is expressed as whitespace, it depends on the platform where you edit this file whether the line ending is a LF (Unix convention), or a CRLF (MS Windows convention) If you want to use DFDL character entities, you must turn on the replaceDFDLEntities="true" feature of the documentPart element. --> <tdml:documentPart type="text"><![CDATA[quantity:250; hardnessRating:967; ]]></tdml:documentPart> <!-- In 'text' both XML character entities, and DFDL's own character entities are interpreted. So here is a NUL terminated string that contains a date with some Japanese Kanji characters. The Japanese characters are expressed using XML numeric character entities. The NUL termination is expressed using a DFDL character entity. In this example one has no choice but to use a DFDL character entity. The NUL character (which has character code zero), is not allowed in XML documents, not even using an XML character entity. So you have to write '%NUL;' or '%#x00;' to express it using DFDL character entities. --> <tdml:documentPart type="text" replaceDFDLEntities="true"><![CDATA[1987年10月日 BCE%NUL;]]></tdml:documentPart> <!-- Type 'byte' means use hexadecimal to specify the data. Freeform whitespace is allowed. Actually, any character that is not a-zA-Z0-9 is ignored. So you can use "." or "-" to separate groups of hex digits if you like. --> <tdml:documentPart type="byte"> 9Abf e4c3 A5-E9-FF-00 </tdml:documentPart> <!-- Type 'bits' allows you to specify individual 0 and 1. Any character other than 0 or 1 is ignored. The number of bits does not have to be a multiple of 8. That is, whole bytes are not required. --> <tdml:documentPart type="bits"> 1.110 0.011 1 First 5 bit fields. </tdml:documentPart> <!-- Type 'file' means the content is a file name where to get the data --> <tdml:documentPart type="file">/some/directory/testData.in.dat</tdml:documentPart> </tdml:document>
Further details on TDML will go in a more detailed guide/page about writing TDML.
If you use the external schema file or external data file capabilities, then of course you need to send those files along with your TDML.
Specifiying the Infoset in XML or an External File
The infoset can be provided either as an inline XML infoset or as a path to an external file by setting the infosetType attribute. If not provided, it defaults to inline XML infoset"
<tdml:infoset> <tdml:dfdlInfoset type="infoset"> <ex:myTestRoot>2013-04-02T14:00:56.000000-05:00</ex:myTestRoot> </tdml:dfdlInfoset> </tdml:infoset> <tdml:infoset> <tdml:dfdlInfoset type="file">/some/directory/testData.in.xml</tdml:dfdlInfoset> </tdml:infoset>
A poor or missing diagnostic message is a bug just as much as a broken feature. So one can also create negative tests, i.e., tests that expect errors.
To do this replace the tdml:infoset element with a tdml:errors element:
<tdml:errors> <tdml:error>Schema Definition Error</tdml:error> <tdml:error>testElementName</tdml:error> </tdml:errors>
Each tdml:error child element contains a sub-string which must be found somewhere in the set of diagnostic messages that come out of the test. The comparison is case-insensitive.
Final detail: In order for a positive test with an Infoset to pass, it must consume all the data. Otherwise the test will not pass, and will fail with a message about 'left over data'.