I recently worked on a project that required manipulating GraphQL documents for queries and mutations. I realized that some techniques I had previously used for JSON and XML could also be helpful when working with GraphQL documents.
GraphQL and JSON are two separate schema files that serve different purposes and are not always used for externalizing or retrieving service data. Unlike JSON, which is primarily used for data storage and exchange, GraphQL documents are utilized specifically for sending queries and mutations to the GraphQL service, and for no other purposes.
To modify a GraphQL document, the following steps need to be followed:
- Traverse through the GraphQL document.
- Identify the relevant GraphQL node or nodes.
- Manipulate the GraphQL node or nodes as required.
- Create a new GraphQL document with the manipulated node or nodes.
- Pass the new GraphQL document to the GraphQL server to execute the query or mutation.
Let's exclude point (5) from our discussion, as it can be accomplished through multiple tools and code snippets.
I found no solution for these steps, so I built an open-source called eXtendGqL. This article will walk you through how you can use eXtendGqL to traverse a GraphQL document, select the relevant nodes, and manipulate the GraphQL document according to your needs.
We'll also play with the code and elaborate on relevant use cases.
eXtendGql is a Java-based library.
POM file
<dependency>
<groupId>com.graphql-java</groupId>
<artifactId>graphql-java</artifactId>
<version>20.4</version>
</dependency>
##Traverse over GraphQL document
Every element in GraphQL, such as document, query, mutation, fragment, inline fragment, directive, etc., is derived from a node with unique attributes and behavior.
eXtendGql uses the observer pattern pattern where the GraphQL document acts as a subject, traverses the entire document and notifies relevant observers on the node and context.
This observer pattern separates the traversal over the GraphQL document from the execution part that the observer consumer code would like to perform.
##Selection of Node
XML has XPath. JSON has JSONPath.
And ... GraphQL document has gXPath
gXPath can be used to navigate through nodes in a GraphQL document. gXPath uses path expressions to select nodes or node on the GraphQL document. Behind the scenes, the gXPath utilizes the traversal module and selects the node according to the required expression.
###gXPath syntax gXPath uses path expressions to select nodes in GraphQL document. The node is selected by following a path or steps.
The most useful path expressions are listed below:
Expression | Description |
---|---|
// | Path prefix: Select all nodes from the root node, and use a slash as a separator between path elements. |
/ | Path prefix: Select the first node from the root node and use a slash as a separator between path elements. The range is not supported when the first node is selected. |
{x:y}/ | Path prefix, Select path node(s), between a range of x path node and y path node (inclusion), use of slash as a separator between path elements. x and y are positive integers. All nodes are selected if no x and y are not set. |
{:y}// | Path prefix, Select path node(s), between a range of first path node result to y, using a slash as a separator between path elements. x and y are positive integers. All nodes are selected if no x and y are not set. |
{x:}/ | Path prefix, Select path node(a), between range of x path node result to the end of path nodes result, use slash as a separator between path elements. x and y are integers. if no x and y are not set, select all path nodes. |
{:}// | Path prefix, Select node(s) from the root node, use of slash as a separator between path elements. |
... | Support of relative path "any" selection e.g. {x:y}//a/b/.../f any can be set anywhere in the gXPath, except at the end of the gXPath, You can set many any as you request, this will help you while selecting node in large GraphQL structure, so you won't be required to mention/build the entire node structure. |
The library also provides an equivalent code named SyntaxPath that provides gXPath expression abilities use of code, mainly used by automation code.
##Transformer The transformer provides the ability to transform (Manipulate) GraphQL document simply. The transformer uses the abilities provided by eXtendGql such as: gXPath, SyntaxPath etc.
The eXtendGql provides the following transform methods:
- Add Children - Add children node to selected GraphQL node or nodes
- Add Sibling - Add sibling node to selected GraphQL node or nodes
- Duplicate - Duplicate selected node by duplication number, multi nodes cannot be duplicated
- Remove Children - Remove selected nodes or node
- Update name - Update selected nodes names or node names, for inline fragments, it will update the typeCondition name
- Update alias value - update field alias value and fragment spread alias value.
Let's start with traversal over the GrqphQL document
Start with creating my observer, let's name it: StringBuilderObserver
The observer will append the GraphQL node to some StringBuilder.
This way, we achieve separation of concern:
- Traverse over the GraphQL document nodes
- Append node values with StringBuilder ....
public class StringBuilderObserver implements TraversalObserver {
private final List<StringBuilderElem> stringBuilderElems = new ArrayList<>();
private final boolean isIgnoreCollection = true;
@Override
public void updateNodeEntry(Node node, Node parentNode, Context context, ObserverAction observerAction) {
String message = "";
DocumentElementType documentElementType = context.getDocumentElementType();
switch (documentElementType) {
case DOCUMENT:
message = MessageFormat.format("Node : {0} || Type : {1}", "Document", documentElementType.name());
break;
case DIRECTIVE:
message = MessageFormat.format("Name : {0} || Type : {1}", ((Directive) node).getName(), documentElementType.name());
break;
case FIELD:
Field field = (Field) node;
message = MessageFormat.format("Name : {0} || Alias : {1} || Type : {2}",
field.getName(),
field.getAlias(),
documentElementType.name());
break;
case OPERATION_DEFINITION:
message = MessageFormat.format("Name : {0} || Type : {1}",
((OperationDefinition) node).getOperation().toString(), documentElementType.name());
break;
case INLINE_FRAGMENT:
message = MessageFormat.format("Node : {0} || Type : {1}", "InlineFragment",
documentElementType.name());
break;
case FRAGMENT_DEFINITION:
message = MessageFormat.format("Name : {0} || Type : {1}",
((FragmentDefinition) node).getName(), documentElementType.name());
break;
case FRAGMENT_SPREAD:
message = MessageFormat.format("Node : {0} || Type : {1}", ((FragmentSpread) node).getName(), documentElementType.name());
break;
case VARIABLE_DEFINITION:
message = MessageFormat.format("Name : {0} || Default Value : {1} || Type : {2}",
((VariableDefinition) node).getName(), ((VariableDefinition) node).getDefaultValue(), documentElementType.name());
break;
case ARGUMENT:
message = MessageFormat.format("Name : {0} || Value : {1} || Type : {2}",
((Argument) node).getName(),
((Argument) node).getValue(),
documentElementType.name());
break;
case ARGUMENTS:
if(isIgnoreCollection) return;
message = MessageFormat.format("Node : {0} || Type : {1}", "Arguments", documentElementType.name());
break;
case SELECTION_SET:
if(isIgnoreCollection) return;
message = MessageFormat.format("Node : {0} || Type : {1}", "SelectionSet", documentElementType.name());
break;
case VARIABLE_DEFINITIONS:
if(isIgnoreCollection) return;
message = MessageFormat.format("Node : {0} || Type : {1}", "VariableDefinitions", documentElementType.name());
break;
case DIRECTIVES:
if(isIgnoreCollection) return;
message = MessageFormat.format("Node : {0} || Type : {1}", "Directives", documentElementType.name());
break;
case DEFINITIONS:
if(isIgnoreCollection) return;
message = MessageFormat.format("Node : {0} || Type : {1}", "Definitions", documentElementType.name());
break;
}
if(Strings.isNullOrEmpty(message)){
return;
}
stringBuilderElems.add(new StringBuilderElem(message, context.getLevel()));
levels.add(context.getLevel());
//spaces++;
}
private final List<Integer> levels = new ArrayList<>();
public String getGqlBrowsedPrintedString() {
return getStringAs(true);
}
private String getStringAs(boolean isIdent) {
int j=0;
StringBuilder stringBuilder = new StringBuilder();
for (StringBuilderElem stringBuilderElem : stringBuilderElems) {
j++;
String spaceStr = "";
if( isIdent) {
for (int i = 0; i < stringBuilderElem.getDepth(); i++) {
spaceStr += " ";
}
stringBuilder.append(spaceStr + stringBuilderElem.getName() + "\n");
}else{
stringBuilder.append( stringBuilderElem.getName() + (j+1<stringBuilderElems.size()? " " : "") );
}
}
return stringBuilder.toString();
}
public String getGqlBrowsedString(){
return getStringAs(false);
}
// int spaces = 0;
@Override
public void updateNodeExit( Node node,Node parentNode, Context context, ObserverAction observerAction) {
}
}
GqlTraversal traversal = new GqlTraversal();
StringBuilderObserver gqlStringBuilderObserver = new StringBuilderObserver();
traversal.getGqlTraversalObservable().addObserver(gqlStringBuilderObserver);
traversal.traverse(file);
System.out.println( gqlStringBuilderObserver.getGqlBrowsedString());
After we saw how the traversal is working, lets digg with some example of gXPath selection GraphQL nodes
gXPath define an expression language as defined above, in addition the expression language contains more terms to familiar with, Element Name and types abbreviations. Why its important, GraphQL document is more than structure that similar to JSON, GraphQL also provide a DSL that exposed by the GraphQL server and GraphQL language. the types and the element name will assist the gXPath to select the exact node or nodes.
Element Names
element_name | Description |
---|---|
type= | Select element by type abbreviate |
name= | Select element by name |
alias= | Select element by alias name |
Available types and abbreviation
Type abbreviate | Description |
---|---|
doc | DOCUMENT |
frag | FRAGMENT_DEFINITION |
direc | DIRECTIVE |
fld | FIELD |
mutation | MUTATION_DEFINITION |
query | OPERATION_DEFINITION |
infrag | INLINE_FRAGMENT |
var | VARIABLE_DEFINITION |
arg | ARGUMENT |
Let's practice the gXPath expression
GraphQL Document
In this document, we have 2 node named: 'name', but of different types: argument and field.
query {
Instrument(Name: "1234") {
Reference {
Name
title
}
}
}
-
Select all nodes (double slash) 'name' which is an argument type
//query/Instrument/name[type=arg]
-
Select first node (single slash) 'name' which is an argument type
/query/Instrument/name[type=arg]
-
//query/.../name[type=arg]
Same as query//query/Instrument/name[type=arg]
-
Select of 'name' node which is field under Reference, reside under Instrument, reside under query
//query/Instrument/Reference/name
-
//query/Instrument/.../name
Same as//query/Instrument/Reference/name
-
//.../name
Same as//query/Instrument/Reference/name
GraphQl Document
query Hero($episode: Episode, $withFriends: Boolean!) {
hero(episode: $episode) {
name
friends @include(if: $withFriends) {
name
}
friends @include(if: $withFriends) {
name
}
friends @include(if: $withFriends) {
name
}
friends @include(if: $withFriends) {
name
}
}
}
-
Select all query nodes named hero
//query[name=hero]
-
Select first query (single slash) node named hero
/query[name=hero]
-
Select all nodes named 'name', reside under friends node
//query[name=hero]/hero/friends/name
-
Select all nodes named 'name' within a range of index 0 and index 2 (inclusion), reside under friends node
{0:2}//query[name=hero]/hero/friends/name
-
Select node named 'name', reside under any node, which reside under hero node
/query[name=hero]/hero/.../name
-
Select $withFriends variable reside directly under the root node named hero
//query[name=hero]/withFriends[type=var]
-
Select include directive first node reside under friends, which reside under root query node named hero
/query[name=hero]/hero/friends/include[type=direc]
-
Select the 'if' argument node, reside under the @include directive
//.../include[type=direc]/if[type=arg]
-
Select episode variable
//.../episode[type=var]
How to use gXPath in the code:
SelectorFacade selectorFacade = new SelectorFacade();
String queryString = Files.readString(file.toPath());
// query { Instrument(id: "1234") }
GqlNodeContext select = selectorFacade.select(queryString, "//query/Instrument / Reference /");
Use of SyntaxPath
String queryString = Files.readString(file.toPath());
SyntaxBuilder eXtendGqlBuilder = new SyntaxBuilder();
eXtendGqlBuilder.appendQuery();
eXtendGqlBuilder.appendField("Instrument");
eXtendGqlBuilder.appendField("Reference");
// query { Instrument(id: "1234") }
GqlNodeContext select = selectorFacade.select(queryString, eXtendGqlBuilder.build());
And finally we will dwell on an example that will illustrate the use of gXPath node selection GraphQL manipulation (AKA transform).
Here a mutation GraphQL document
mutation CreateReviewForEpisode($ep: Episode!, $review: ReviewInput!) {
createReview(episode: $ep, review: $review) {
stars
commentary
stars
commentary
}
}
The following java code will select and transform mutation document,
String queryString = Files.readString(sourceFile.toPath());
TransformBuilder transformBuilder = new TransformBuilder();
transformBuilder.addChildrenNode("//mutation[name=CreateReviewForEpisode]/createReview/stars",new Field("child_of_stars"))
.addSiblingNode("//mutation[name=CreateReviewForEpisode]/createReview/stars",new Field("sibling_of_stars"))
.updateNodeName("//mutation[name=CreateReviewForEpisode]/createReview/stars","star_new_name")
.removeNode("//mutation[name=CreateReviewForEpisode]/createReview/commentary")
.duplicateNode("//mutation[name=CreateReviewForEpisode]/createReview/sibling_of_stars", 10);
TransformExecutor transformExecutor = new TransformExecutor(transformBuilder);
RawPayload rawPayload = new RawPayload();
rawPayload.setQueryValue(queryString);
RawPayload executeRawPayload = transformExecutor.execute(rawPayload);
Description:
add new children node named: child_of_stars under gXPath:
//mutation[name=CreateReviewForEpisode]/createReview/stars
Add new sibling node named: sibling_of_stars under gXPath selected node:
//mutation[name=CreateReviewForEpisode]/createReview/stars
Set new node name: star_new_name value to the selected node by gXPath:
//mutation[name=CreateReviewForEpisode]/createReview/stars
Remove selected node by gXPath
//mutation[name=CreateReviewForEpisode]/createReview/commentary
Duplicate selected gXPath node 10 times
//mutation[name=CreateReviewForEpisode]/createReview/sibling_of_stars
Use TransformBuilder to build the transform plan, with selected node use of gXPath and the command to execute. The transform plan is load to the TransformExecutor, with the GraphQL payload.
The execution will result in a new GraphQL document,
mutation CreateReviewForEpisode($ep: Episode!, $review: ReviewInput!) {
createReview(episode: $ep, review: $review) {
sibling_of_stars
star_new_name {
child_of_stars
}
star_new_name {
child_of_stars
}
sibling_of_stars
sibling_of_stars
sibling_of_stars
sibling_of_stars
sibling_of_stars
sibling_of_stars
sibling_of_stars
sibling_of_stars
sibling_of_stars
sibling_of_stars
}
}
Last example, will demonstrate the manipulation of directive from include to exclude
Here the query GraphQL document,
query Hero($episode: Episode, $withFriends: Boolean!) {
hero(episode: $episode) {
name
friends **@include**(if: $withFriends) {
name
}
}
}
Here the code:
String queryString = Files.readString(file.toPath());
// query { Instrument(id: "1234") }
GqlNodeContext includeDirectiveNode = selectorFacade.selectSingle(queryString, "//query[name=hero]/hero/friends/include[type=direc]");
assertNotNull(includeDirectiveNode);
assertTrue(includeDirectiveNode.getType().equals(DocumentElementType.DIRECTIVE));
System.out.println("\nBefore manipulation:\n\n" + queryString);
// Node newNode = new Field("new_name");
Node excludeDirectiveNode = TransformUtils.updateNodeName(includeDirectiveNode, "exclude");
String newGqlValue = eXtendGqlWriter.writeToString(excludeDirectiveNode);
System.out.println("\nAfter manipulation:\n\n" + newGqlValue);
GqlNodeContext excludeUpdateNode = selectorFacade.selectSingle(newGqlValue, "//query[name=hero]/hero/friends/exclude[type=direc]");
assertTrue(excludeUpdateNode.getType().equals(DocumentElementType.DIRECTIVE));
query Hero($episode: Episode, $withFriends: Boolean!) {
hero(episode: $episode) {
name
friends @exclude(if: $withFriends) {
name
}
}
}
The eXtendGql can be used during base code while the developer required to enrich GraphQL document with more fields, while querying server for data, or during manipulation of data in server, so the code can articulate the relevant fields to manipulate in the service side. The eXtendGql can also be utilize during integration or E2E testing, generating of synthetic GraphQL data, result with high velocity and managed solution.
I elaborate the following use cases with more details:
Synthetic GraphQL document creation Testing, mostly integration testing part will demands the ability to query GraphQL server with different queries and mutations.
Of course, the developer can maintain large lists of example files to send to the server or to find and replace the relevant string in GraphQL document, but it is a cumbersome solution, hard to maintain etc.
eXtendGqL gives you the ability to manipulate the query or the mutation with ease and use configuration-wise to list your plan of action (gXPath, Transform Commands, and Argument to execute], and execute the plan:
Configuration
Plan
steps
step 1
gXPath (String)
transform_commands
transform_command
command
argument_object_definition
step n
gXPath (String)
transform_commands
transform_command
command
argument_object_definition
origin_file_to_manipulate
Configuration config = read_configuration_plan(plan_file);
config_verification()
build_plan -> ... use of TransformBuilder
new_graphql_document = execute_plan -> ... use of TransformExecuter
eXtendGqL gives you versatility and the ability to produce synthetic GraphQL data and verify the integrity of a GraphQL service.
Articulate GraphQL document on-the-fly
Sometimes, you may need to dynamically build a query or mutation based on business logic or configuration and send it to the GraphQL server. The eXtendGql library can assist with this.
The eXtendGql library only allows manipulation of the GraphQL skeleton file, not its creation.
The developer can create the skeleton GraphQL file, store the file in resource folder. skeleton file, means file with structure but without field only. with the of syntaxPath, eXtendGqL can assemble the gXpath and set the plan strategy on the fly, then use of TransformBuilder to build the plan, then use the TransformExecutor to run the plan.