Message templates

You're viewing Apigee Edge documentation.
Go to the Apigee X documentation. info

This topic discusses how to use message templates in API proxies and provides a function reference.

What is a message template?

A message template permits you to perform variable string substitution in certain policy and TargetEndpoint elements. This feature, where supported, lets you populate strings dynamically when a proxy executes.

You can include any combination of flow variable references and literal text in a message template. Flow variable names must be enclosed in curly braces, while any text not in curly braces is output as literal text.

See also Where can you use message templates?

Example

For example, the Assign Message policy lets you use a message template within in the <Payload> element:

<AssignMessage name="set-dynamic-content">
  <AssignTo createNew="false" type="response"></AssignTo>
  <Set>
    <Payload contentType="application/json">
      {"name":"Alert", "message":"You entered an invalid username: {user.name}"}
    </Payload>
  </Set>
  <IgnoreUnresolvedVariables>false</IgnoreUnresolvedVariables>
</AssignMessage>

In the above example, the value of flow variable user.name (in curly braces) will be evaluated and substituted into the payload string at runtime. So, for example, if user.name=jdoe, then the resulting message output in the payload will be: You entered an invalid username: jdoe. If the variable cannot be resolved, then an empty string is output.

Example

When a quota is exceeded, it is a good practice to return a meaningful message to the caller. This pattern is commonly used with a "fault rule" to provide output to give the caller information about the quota violation. In the following Assign Message policy, message templates are used to populate quota information dynamically in several XML elements:

<AssignMessage name='AM-QuotaViolationMessage'>
  <Description>message for quota exceeded</Description>
  <IgnoreUnresolvedVariables>false</IgnoreUnresolvedVariables>
  <Set>
    <Headers>
      <Header name='X-Quota-Reset'>{ratelimit.Quota-1.expiry.time}</Header>
      <Header name='X-Quota-Allowed'>{ratelimit.Quota-1.allowed.count}</Header>
      <Header name='X-Quota-Available'>{ratelimit.Quota-1.available.count}</Header>
    </Headers>
    <Payload contentType='application/json'>{
  "error" : {
    "message" : "you have exceeded your quota",
    "clientId" : "{request.queryparam.apikey}"
  }
}
    </Payload>
    <StatusCode>429</StatusCode>
    <ReasonPhrase>Quota Exceeded</ReasonPhrase>
  </Set>
</AssignMessage>

In the AssignMessage policy, the following elements in the <Set> element support message templating:

• Header
• QueryParam
• FormParam
• PayLoad
• Version
• Verb
• Path
• StatusCode
• ReasonPhrase

Again, note that flow variables in a message template must be enclosed in curly braces.

When this policy executes:

• The Header elements receive values of the specified flow variables.
• The Payload includes a mix of literal text and variables (the client_id is populated dynamically).
• The StatusCode and ReasonPhrase only include literal text; however, these elements also support message templating if you wish to use it.

Example

In a proxy TargetEndpoint definition, child-elements of <SSLInfo> support message templating. Following the same pattern used in policies, the flow variables in curly braces are replaced when the proxy executes.

<TargetEndpoint name="default">
  …
  <HTTPTargetConnection>
    <SSLInfo>
        <Enabled>{myvars.ssl.enabled}</Enabled>
        <ClientAuthEnabled>{myvars.ssl.client.auth.enabled}</ClientAuthEnabled>
        <KeyStore>{myvars.ssl.keystore}</KeyStore>
        <KeyAlias>{myvars.ssl.keyAlias}</KeyAlias>
        <TrustStore>{myvars.ssl.trustStore}</TrustStore>
    </SSLInfo>

  </HTTPTargetConnection>
  …
</TargetEndpoint>

Where can you use message templates?

Message templates are supported in several policies as well as certain elements used in the TargetEndpoint configuration.

Policies that accept message templates

TargetEndpoint elements that accept message templates

Message template syntax

This section explains the rules you must follow to use message templates.

Use curly braces to denote variables

Enclose variable names in curly braces { }. If the variable does not exist, an empty string is returned in the output; however, you can specify default values in message templates (values that are substituted if the variable is unresolved). See Setting default values in message templates.

Note that enclosing the entire message template string in quotes is permitted, but optional. For example, the following two message templates are equivalent:

<Set>
    <Headers>
        <Header name="x-h1">"Hello {user.name}"</Header>
        <Header name="x-h1">Hello {user.name}</Header>
    </Headers>
</Set>

Setting default values in message templates

If a templated variable can't be resolved, Edge substitutes an empty string. However, you can specify a default value as follows:

<Header name="x-h1">Test message. id = {request.header.id:Unknown}</Header>

In the above sample if the variable request.header.id cannot be resolved then its value is replaced by Unknown. For example:

Test message. id = Unknown

Spaces are not allowed in function expressions

Spaces are not allowed anywhere in message template function expressions. For example:

Allowed:

{substring(alpha,0,4)}
{createUuid()}
{randomLong(10)}

Not Allowed:

{substring( alpha, 0, 4 )}
{ createUuid( ) }
{randomLong( 10 )}

Legacy syntax for JSON payloads

In Edge versions before Cloud release 16.08.17, you could not use curly braces to denote variable references within JSON payloads. In those older versions, you needed to use the variablePrefix and variableSuffix attributes to specify delimiter characters, and use those to wrap variable names, like so:

<Set>
  <Payload contentType="application/json" variablePrefix="@" variableSuffix="#">
    {"name":"foo", "type":"@variable_name#"}
  </Payload>
</Set>

Although Apigee recommends that you use the newer curly-brace syntax, the older syntax still works.

Using message template functions

Edge provides a set of functions that you can use within message templates to escape, encode, hash, and format string variables.

The message template functions are described in detail in Message template function reference.

Example: toLowerCase()

Use the built-in toLowerCase() function to transform a string variable to lowercase:

<AssignMessage name="AM-Set-Custom-Response">
    <AssignTo createNew="false" type="response"/>
    <IgnoreUnresolvedVariables>true</IgnoreUnresolvedVariables>
    <Set>
        <Headers>
            <Header name="x-h1">Test header: {toLowerCase(foo.bar:FOO)}</Header>
        </Headers>
    </Set>
</AssignMessage>

If the foo.bar flow variable resolves, then its characters will be all lowercase. If foo.bar is unresolved then the default value FOO is substituted and converted to lowercase characters. For example:

Test header: foo

Example: escapeJSON()

Here's an interesting use case: Let's say your backend app returns a JSON response that contains valid escape characters. For example:

{
      "code": "INVALID",
      "user_message": "Invalid value for \"logonId\" check your input."
}

Then, let's say you want to return this message to the client caller in a custom payload. The usual way to do this is to extract the message from the target response payload and use Assign Message to add it to a custom proxy response (that is, send it back to the client).

Here's the Extract Variables policy that extracts the user_message information into a variable called standard.systemMessage:

<ExtractVariables name="EV-BackendErrorResponse">
    <DisplayName>EV-BackendErrorResponse</DisplayName>
    <JSONPayload>
        <Variable name="standard.systemMessage">
            <JSONPath>$.user_message</JSONPath>
        </Variable>
    </JSONPayload>
</ExtractVariables>

Now, here's a perfectly valid Assign Message policy that adds the extracted variable to the response payload (the proxy response):

<AssignMessage name="AM-SetStandardFaultResponse">
    <DisplayName>AM-SetStandardFaultResponse</DisplayName>
    <Set>
        <Payload contentType="application/json">
           {
              "systemMessage": "{standard.systemMessage}"
           }
        </Payload>
    </Set>
    <IgnoreUnresolvedVariables>true</IgnoreUnresolvedVariables>
    <AssignTo createNew="false" transport="http" type="response"/>
</AssignMessage>

Unfortunately, there's a problem. The Extract Variables policy removed the escaped quotation characters around part of the message. This means that the response returned to the client is invalid JSON. This is clearly not what you intended!

{
    "systemMessage": "Invalid value for "logonId" check your input."
}

To work around this problem, you can modify the Assign Message policy to use a message template function that escapes the quotes within the JSON for you. This function, escapeJSON(), escapes any quotes or other special characters that occur within a JSON expression:

<AssignMessage name="AM-SetStandardFaultResponse">
    <DisplayName>AM-SetStandardFaultResponse</DisplayName>
    <Set>
        <Payload contentType="application/json">
           {
              "systemMessage": "{escapeJSON(standard.systemMessage)}"
           }
        </Payload>
    </Set>
    <IgnoreUnresolvedVariables>true</IgnoreUnresolvedVariables>
    <AssignTo createNew="false" transport="http" type="response"/>
</AssignMessage>

The function escapes the embedded quotes, resulting in valid JSON, which is exactly what you wanted:

{
      "systemMessage": "Invalid value for \"logonId\" check your input.",
}

A message template is a dynamic string substitution feature that you can use in certain policies and in TargetEndpoint definitions. Message template functions let you perform useful operations such as hashing, string manipulation, character escaping, and others within a message template.

For example, in the following AssignMessage policy, the function toLowerCase() is used in a message template:

<AssignMessage name="AM-Set-Custom-Response">
    <AssignTo createNew="false" type="response"/>
    <IgnoreUnresolvedVariables>true</IgnoreUnresolvedVariables>
    <Set>
       <Headers>
         <Header name="x-h1">Test header: {Hello, toLowerCase(user.name)}</Header>
       </Headers>
    </Set>
</AssignMessage>

This topic describes the message template functions, their arguments, and outputs. This topic assumes that you are familiar with message templates and the contexts in which they are used.

Hash functions

Compute a hash value and return the string representation of that hash.

Hexadecimal hash functions

Compute a hash value and return the string representation of that hash as a hexadecimal number.

Syntax

Arguments

string - The Hash functions take a single string argument upon which the hash algorithm is computed. The argument can be a literal string or a string flow variable.

Examples

Function call:

sha256Hex('abc')

Result:

ba7816bf8f01cfea414140de5dae2223b00361a396177a9cb410ff61f20015ad

Function call:

var str = 'abc';
sha256Hex(str)

Result:

ba7816bf8f01cfea414140de5dae2223b00361a396177a9cb410ff61f20015ad

Base64 hash functions

Compute a hash value and return the string representation of that hash as a Base64 encoded value.

Syntax

Arguments

string - The Hash functions take a single string argument upon which the hash algorithm is computed. The argument can be a literal string or a string flow variable.

Examples

Function call:

sha256Base64('abc')

Result:

ungWv48Bz+pBQUDeXa4iI7ADYaOWF3qctBD/YfIAFa0=

Function call:

var str = 'abc';
sha256Base64(str)

Result:

ungWv48Bz+pBQUDeXa4iI7ADYaOWF3qctBD/YfIAFa0=

String functions

Perform operations on strings within a message template.

Base64 encoding functions

Encode and decode strings using the Base64 encoding scheme.

Syntax

Arguments

string - The string to encode or decode. Can be a literal string or a string flow variable.

Example

<AssignMessage name="AM-Set-Custom-Response">
    <AssignTo createNew="false" type="response"/>
    <IgnoreUnresolvedVariables>true</IgnoreUnresolvedVariables>
    <Set>
       <Headers>
         <Header name="x-h1">Hello, {decodeBase64('d29ybGQK')}</Header>
       </Headers>
    </Set>
</AssignMessage>

Case conversion functions

Convert a string to all uppercase or all lowercase letters.

Syntax

Arguments

string - The string to convert. Can be a literal string or a string flow variable.

Example

<AssignMessage name="AM-Set-Custom-Response">
    <AssignTo createNew="false" type="response"/>
    <IgnoreUnresolvedVariables>true</IgnoreUnresolvedVariables>
    <Set>
       <Headers>
         <Header name="x-h1">Hello, {toLowerCase(user.name)}</Header>
       </Headers>
    </Set>
</AssignMessage>

Substring function

Returns the characters between the starting and ending index of the specified string.

Syntax

substring(str,start_index,end_index)

Arguments

• str - A literal string or string flow variable.
• start_index - The starting index into the string.
• end_index - (Optional) The ending index into the string. If not supplied, ending index is the end of the string.

Examples

For the following examples, suppose that these flow variables exist:

Here are results of function calls that use these variables:

Replace All function

Applies a regular expression to a string and for any matches, replaces the match with a replacement value.

Syntax

replaceAll(string,regex,value)

Arguments

• string - A literal string or string flow variable in which to make replacements.
• regex - A regular expression.
• value - The value to replace all regex matches within the string.

Examples

For the following examples, suppose these flow variables exist:

Here are results of function calls that use these variables:

Replace First function

Replaces only the first occurrence of the specified regular expression match in the string.

Syntax

replaceFirst(string,regex,value)

Arguments

• string - A literal string or string flow variable in which to make replacements.
• regex - A regular expression.
• value - The value to replace regex matches within the string.

Character escape and encoding functions

Functions that escape or encode special characters in a string.

Syntax

Arguments

string - The string to escape. Can be a literal string or a string flow variable.

Usage notes

XML 1.1 can represent certain control characters, but it cannot represent the null byte or unpaired Unicode surrogate codepoints, even after escaping. The escapeXML11() function removes characters that do not fit in the following ranges:

[#x1-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]

The escapeXML11() function escapes characters in the following ranges:

[#x1-#x8] | [#xB-#xC] | [#xE-#x1F] | [#x7F-#x84] | [#x86-#x9F]

Examples

Assume a flow variable called food exists with this value: "bread" & "butter". Then, the function:

{escapeHTML(food)}

results in:

Time format functions

Return a string representation of the time, formatted in the local time zone, or in UTC.

Syntax

Arguments

• format - A date/time format string. Can be a literal string or a string variable.
• str - A string or string flow variable containing a time value. The value can be in seconds-since-epoch or milliseconds-since-epoch for timeFormatMs.

Examples

Assume the following values and assume the local timezone is Pacific:

• epoch_time_ms = 1494390266000
• epoch_time = 1494390266
• fmt1 = yyyy-MM-dd
• fmt2 = yyyy-MM-dd HH-mm-ss
• fmt3 = yyyyMMddHHmmss

The functions return the following results:

Function	Output
timeFormatMs(fmt1,epoch_time_ms)	2017-05-09
timeFormat(fmt1,epoch_time)	2017-05-09
timeFormat(fmt2,epoch_time)	2017-05-09 21:24:26
timeFormat(fmt3,epoch_time)	20170509212426
timeFormatUTC(fmt1,epoch_time)	2017-05-10
timeFormatUTC(fmt2,epoch_time)	2017-05-10 04:24:26
timeFormatUTC(fmt3,epoch_time)	20170510042426

HMAC calculation functions

The HMAC calculation functions provide an alternative to using the HMAC policy to compute an HMAC. The functions are handy when performing a cascaded HMAC calculation, as when the output of one HMAC is used as the key for a second HMAC.

Syntax

Function	Description
hmacSha224(key,valueToSign[,keyencoding[,outputencoding]])	Computes an HMAC with the SHA-224 hash function.
hmacSha256(key,valueToSign[,keyencoding[,outputencoding]])	Encodes an HMAC with the SHA-256 hash function.
hmacSha384(key,valueToSign[,keyencoding[,outputencoding]])	Encodes an HMAC with the SHA-384 hash function.
hmacSha512(key,valueToSign[,keyencoding[,outputencoding]])	Encodes an HMAC with the SHA-512 hash function.
hmacMd5(key,valueToSign[,keyencoding[,outputencoding]])	Encodes an HMAC with the MD5 hash function.
hmacSha1(key, valueToSign [,keyencoding[,outputencoding]])	Encodes an HMAC with the SHA-1 encryption algorithm.

Arguments

• key - (Required) Specifies the secret key, encoded as a string, used to compute the HMAC.
• valueToSign - (Required) Specifies the message to sign. It should be a string.
• keyencoding - (Optional) The secret key string will be decoded according to this specified encoding. Valid values: hex, base16, base64, utf-8. Default: utf-8
• outputencoding - (Optional) Specifies the encoding algorithm to use for the output. Valid values: hex, base16, base64. The values are case insensitive; hex and base16 are synonyms. Default: base64

Examples

This example uses the AssignMessage policy to compute an HMAC-256 and assign it to a flow variable:

<AssignMessage name='AM-HMAC-1'>
  <AssignVariable>
    <Name>valueToSign</Name>
    <Template>{request.header.apikey}.{request.header.date}</Template>
  </AssignVariable>
  <AssignVariable>
    <Name>hmac_value</Name>
    <Template>{hmacSha256(private.secretkey,valueToSign)}</Template>
  </AssignVariable>
</AssignMessage>

This example illustrates how to generate a cascading HMAC that can be used with the AWS Signature v4 signing process. The example uses the AssignMessage policy to generate the five levels of cascaded HMAC used to calculate a signature for AWS Signature v4:

<AssignMessage name='AM-HMAC-AWS-1'>
  <!-- 1 -->
  <AssignVariable>
    <Name>DateValue</Name>
    <Template>{timeFormatUTCMs('yyyyMMdd',system.timestamp)}</Template>
  </AssignVariable>
  <!-- 2 -->
  <AssignVariable>
    <Name>FirstKey</Name>
    <Template>AWS4{private.secret_aws_access_key}</Template>
  </AssignVariable>
  <!-- 3 -->
  <AssignVariable>
    <Name>DateKey</Name>
    <Template>{hmacSha256(FirstKey,DateValue,'utf-8','base16')}</Template>
  </AssignVariable>
  <!-- 4 -->
  <AssignVariable>
    <Name>DateRegionKey</Name>
    <Template>{hmacSha256(DateKey,aws_region,'base16','base16')}</Template>
  </AssignVariable>
  <!-- 5 -->
  <AssignVariable>
    <Name>DateRegionServiceKey</Name>
    <Template>{hmacSha256(DateRegionKey,aws_service,'base16','base16')}</Template>
  </AssignVariable>
  <!-- 6 -->
  <AssignVariable>
    <Name>SigningKey</Name>
    <Template>{hmacSha256(DateRegionServiceKey,'aws4_request','base16','base16')}</Template>
  </AssignVariable>
  <!-- 7 -->
  <AssignVariable>
    <Name>aws4_hmac_value</Name>
    <Template>{hmacSha256(SigningKey,stringToSign,'base16','base16')}</Template>
  </AssignVariable>
</AssignMessage>

Other functions

Create UUID function

Generates and returns a UUID.

Syntax

createUuid()

Arguments

None.

Example

{createUuid()}

Sample result:

ec3ca9be-d1e1-4ef4-aee4-4a58f3130db8

Random Long Generator function

Returns a random long integer.

Syntax

randomLong(args)

Arguments

• If no arguments are specified, the function returns a random long integer, as computed by the Java SecureRandom class.
• If one argument is present, it is treated as the minimum value of the computation.
• If a second argument is present, it is treated as the maximum value of the computation.

Example

{random()}

results in something like this:

5211338197474042880

Regex text generator

Generate a text string that matches a given regular expression.

Syntax

xeger(regex)

Argument

regex - A regular expression.

Example

This example generates a seven digit string with no zeros:

xeger('[1-9]{7}')

Example result:

9857253

Null-coalescing function

The firstnonnull() function returns the value of the left-most, non-null argument.

Syntax

firstnonnull(var1,varn)

Argument

var1 - A context variable.

var_n - One or more context variables. You can set the right-most argument to a string to provide a fallback value (a value that will be set if none of the left-hand arguments are set).

Examples

The following table illustrates how to use the function:

Template	Var1	Var2	Var3	Result
{firstnonnull(var1,var2)}	Not set	foo	N/A	foo
{firstnonnull(var1,var2)}	foo	bar	N/A	foo
{firstnonnull(var1,var2)}	foo	Not set	N/A	foo
{firstnonnull(var1,var2,var3)}	foo	bar	baz	foo
{firstnonnull(var1,var2,var3)}	Not set	bar	baz	bar
{firstnonnull(var1,var2,var3)}	Not set	Not set	baz	baz
{firstnonnull(var1,var2,var3)}	Not set	Not set	Not set	null
{firstnonnull(var1)}	Not set	N/A	N/A	null
{firstnonnull(var1)}	foo	N/A	N/A	foo
{firstnonnull(var1,var2)}	""	bar	N/A	""
{firstnonnull(var1,var2,'fallback value')}	null	null	fallback value	fallback value

XPath function

Applies an XPath expression to an XML variable.

Syntax

xpath(xpath_expression,xml_string,[datatype])

Arguments

xpath_expression - An XPath expression.

xml_string - A flow variable or string containing XML.

datatype - (Optional) Specifies the desired return type of the query. It can be nodeset, node, number, boolean, string. It defaults to nodeset. The default is usually the right choice.

Example 1

Suppose these context variables define an XML string and an XPath expression:

xml = "<tag><tagid>250397</tagid><readerid>1</readerid><rssi>74</rssi><date>2019/06/15</date></tag>"
xpath = "/tag/tagid"

And the xpath() function is used in an AssignMessage policy, as follows:

<AssignMessage>
  <AssignVariable>
    <Name>extracted_tag</Name>
    <Template>{xpath(xpath,xml)}</Template>
  </AssignVariable>
</AssignMessage><

The function returns the value <tagid>250397</tagid>. This value is placed in the context variable called extracted_tag.

Example 2

If you only want the value of the node, use the text() function, as follows:

<AssignMessage>
  <AssignVariable>
    <Name>extracted_tag</Name>
    <Template>{xpath('/tag/tagid/text()',xml)}</Template>
  </AssignVariable>
</AssignMessage>

As a result of this operation, the context variable extracted_tag is set to 250397

If multiple nodes are selected, then the result of xpath() is all of the values of the selection, concatenated with a comma.

Example 3: XML namespaces

To specify a namespace, append additional parameters, each one a string that looks like prefix:namespaceuri. For example, an xpath() function that selects the child element of a SOAP body might be like this:

<AssignMessage>
  <AssignVariable>
    <Name>soapns</Name>
    <Value>soap:http://schemas.xmlsoap.org/soap/envelope/</Value>
  </AssignVariable>
  <AssignVariable>
    <Name>xpathexpression</Name>
    <Value>/soap:Envelope/soap:Body/*</Value>
  </AssignVariable>
  <AssignVariable>
    <Name>extracted_element</Name>
    <Template>{xpath(xpathexpression,xml,soapns)}</Template>
  </AssignVariable>
</AssignMessage>

For additional namespaces, you can add up to 10 additional parameters to the xpath() function.

You can specify a simple XPath expression as a string surrounded by single quotes:

{xpath('/tag/tagid/text()',xml)}

If the XPath expression includes namespace prefixes (and colons) then you need to assign that XPath expression to a variable and specify the variable name rather than the expression directly.

{xpath(xpathexpression,xml,ns1)}

Example 4: Specifying a desired return type

The optional third parameter passed to the xpath() function specifies the desired return type of the query.

Some XPath queries can return numeric or boolean values. For example the count() function returns a number. This is a valid XPath query:

count(//Record/Fields/Pair)

This valid query returns a boolean:

count(//Record/Fields/Pair)>0

In those cases invoke the xpath() function with a third parameter specifying that type:

{xpath(expression,xml,'number')}
{xpath(expression,xml,'boolean')}

If the third parameter contains a colon, then it is interpreted as a namespace argument. If not, then it is treated as the desired return type. In this case, if the third parameter is not one of the valid values (ignoring case), the xpath() function defaults to returning a nodeset.

JSON Path function

Applies a JSON Path expression to a JSON variable.

Syntax

jsonPath(json-path,json-var,want-array)

Arguments

• (Required) json-path: (String) A JSON Path expression.
• (Required) json-var: (String) A flow variable or string containing JSON.
• (Optional) want-array: (String) If this parameter is set to 'true' and if the result set is an array, then all array elements are returned. If set to any other value or if this parameter is omitted, then only the zeroth element of a result set array is returned. If the result set is not an array, then this third parameter, if present, is ignored.

Example 1

If this is the message template:

The address is {jsonPath($.results[?(@.name == 'Mae West')].address.line1,the_json_variable)}

and the_json_variable contains:

{
  "results" : [
    {
      "address" : {
        "line1" : "18250 142ND AV NE",
        "city" : "Woodinville",
        "state" : "Washington",
        "zip" : "98072"
      },
      "name" : "Fred Meyer"
    },
    {
      "address" : {
        "line1" : "1060 West Addison Street",
        "city" : "Chicago",
        "state" : "Illinois",
        "zip" : "60613"
      },
      "name" : "Mae West"
    }
  ]
} 

The result of the function is:

The address is 1060 West Addison Street

Note that in this case, the result set is a single element (not an array of elements). If the result set were an array, then only the zeroth element of the array would be returned. To return the full array, call the function with 'true' as the third parameter, as shown in the next example.

Example 2

If this is the message template:

{jsonPath($.config.quota[?(@.operation=='ManageOrder')].appname,the_json_variable,'true')}

and the_json_variable contains:

{
  "results" : [
     {
      "config": {
        "quota": [
          {
            "appname": "A",
            "operation": "ManageOrder",
            "value": "900"
          },
          {
            "appname": "B",
            "operation": "ManageOrder",
            "value": "1000"
          },
          {
            "appname": "B",
            "operation": "SubmitOrder",
            "value": "800"
          }
        ]
      }
    }
  ]
} 

The result of the function is:

['A','B']