roadmap
- 01
- 02
- 03
- 04
- 05
- 06
- 07
- 08
Prepare for Karate interview questions on testing HTTP APIs in feature files with no step definitions to write. Covers payload matching, reusable calls, config, the mock server and parallel runs.
177 of 177 questions in your plan: 0 known, 0 in progress, 0 due, 0 skipped, 177 new. 0 cut, outside your plan.
Karate usually comes up in the API automation round of a QA or SDET loop, and sometimes as a take-home: a small suite against a public endpoint, read afterwards line by line. Because a feature file reads almost like prose, interviewers test whether you can predict what it actually executes. Which step puts the request on the wire, and what does `status` do after it? Which settings still apply to the second call in a scenario once `path` has been reset? Why does `match response == expected` fail when the reply carries one extra key, while `contains` passes? Candidates who have only copied examples answer with the keyword. Stronger candidates answer with the rule behind it: a server-generated `id` is handled with a marker such as `#notnull`, not by deleting it from the expected payload; a login placed in a `Background` runs again before every scenario unless `callonce` caches it; a called feature shares or isolates its variables depending on whether you assign the result. Senior rounds move to the edges of a whole suite: what a stateful mock does when two requests arrive at once, why a CI job reports no tests although the run passed, and what a team gives up in compile-time checking when it trades a Java library for feature files.
HTTP comes first because nearly every Karate keyword is an HTTP concept with a short name: the request [method and the status codes](/topics/proto-http-methods-status) it can return, and [how an API contract picks them](/topics/proto-rest-http-verbs-status-codes). The plan then turns to the file itself. [Gherkin's Background and outlines](/topics/found-bdd-gherkin) and [what a step definition is](/topics/found-bdd-executable-specs) come before Karate's own syntax, since what Karate removes only makes sense once you know the layer it removes; [polling for readiness](/topics/found-test-automation-waiting-sync) sits beside them because waiting is a built-in keyword here, not a helper you write. With that ground, [glue-free execution](/topics/dev-karate-glue) shows how a line becomes an action. Karate compares JSON and XML natively, so a short stretch on [JSON's value types](/topics/proto-json-syntax-types), [XPath](/topics/proto-xml-xpath-xslt) and [XML namespaces](/topics/proto-xml-namespaces) precedes the largest product module. Assertion follows in two steps: first [what an API-level case should assert](/topics/found-test-automation-api-case-design), then [payload matching](/topics/dev-karate-assert), whose [fuzzy markers and predicates](/topics/dev-karate-assert-tolerance) are how that judgement is written down. The second half moves from one scenario to a suite. [Configuration precedence](/topics/found-test-automation-framework-arch-config-precedence) and [choosing a target environment by name](/topics/found-test-automation-framework-arch-env-resolution) are the decisions; [reuse and defaults](/topics/dev-karate-reuse) is Karate's mechanism for them, from called features to the config file chain. [Service virtualization](/topics/found-test-data-environments-service-virtualization) then frames [mock mode](/topics/dev-karate-mock), where the same feature-file format answers requests instead of sending them. The plan closes on running at scale: [what parallel workers contend for](/topics/found-test-automation-framework-arch-parallel-isolation), then [the Java runner and the adoption argument](/topics/dev-karate-fit).
Coming from Postman, the jump is smaller than it looks: the request keywords and `match` carry most of the junior questions, so spend your first sessions there and narrow level to the range you are applying for. Coming from Cucumber, set aside the instinct to look for glue code; the opening scenario questions are mostly about what Karate does without it. If you already write REST Assured suites, go straight to reuse, mocks and the trade-off module, where the questions turn to judgement, and raise importance to must know when the interview is days away. Senior candidates should expect the mock under concurrent requests, once-per-run caching across parallel threads and the CI wiring to fill most of the conversation. This plan teaches Karate alone. Postman collections, REST Assured, standalone stub servers and consumer-driven contracts sit together in the [API Testing](/roadmaps/api-testing) roadmap, which places Karate among them. Driving a browser with Karate is not part of this plan, and how a load model is designed and read belongs to the [Gatling](/roadmaps/gatling) roadmap; the [QA Engineer](/roadmaps/qa-engineer) role carries the rest of the loop.
Karate usually comes up in the API automation round of a QA or SDET loop, and sometimes as a take-home: a small suite against a public endpoint, read afterwards line by line. Because a feature file reads almost like prose, interviewers test whether you can predict what it actually executes. Which step puts the request on the wire, and what does status do after it? Which settings still apply to the second call in a scenario once path has been reset? Why does match response == expected fail when the reply carries one extra key, while contains passes?
Candidates who have only copied examples answer with the keyword. Stronger candidates answer with the rule behind it: a server-generated id is handled with a marker such as #notnull, not by deleting it from the expected payload; a login placed in a Background runs again before every scenario unless callonce caches it; a called feature shares or isolates its variables depending on whether you assign the result. Senior rounds move to the edges of a whole suite: what a stateful mock does when two requests arrive at once, why a CI job reports no tests although the run passed, and what a team gives up in compile-time checking when it trades a Java library for feature files.
HTTP comes first because nearly every Karate keyword is an HTTP concept with a short name: the request method and the status codes it can return, and how an API contract picks them. The plan then turns to the file itself. Gherkin's Background and outlines and what a step definition is come before Karate's own syntax, since what Karate removes only makes sense once you know the layer it removes; polling for readiness sits beside them because waiting is a built-in keyword here, not a helper you write. With that ground, glue-free execution shows how a line becomes an action.
Karate compares JSON and XML natively, so a short stretch on JSON's value types, XPath and XML namespaces precedes the largest product module. Assertion follows in two steps: first what an API-level case should assert, then payload matching, whose fuzzy markers and predicates are how that judgement is written down.
The second half moves from one scenario to a suite. Configuration precedence and choosing a target environment by name are the decisions; reuse and defaults is Karate's mechanism for them, from called features to the config file chain. Service virtualization then frames mock mode, where the same feature-file format answers requests instead of sending them. The plan closes on running at scale: what parallel workers contend for, then the Java runner and the adoption argument.
Coming from Postman, the jump is smaller than it looks: the request keywords and match carry most of the junior questions, so spend your first sessions there and narrow level to the range you are applying for. Coming from Cucumber, set aside the instinct to look for glue code; the opening scenario questions are mostly about what Karate does without it. If you already write REST Assured suites, go straight to reuse, mocks and the trade-off module, where the questions turn to judgement, and raise importance to must know when the interview is days away.
Senior candidates should expect the mock under concurrent requests, once-per-run caching across parallel threads and the CI wiring to fill most of the conversation.
This plan teaches Karate alone. Postman collections, REST Assured, standalone stub servers and consumer-driven contracts sit together in the API Testing roadmap, which places Karate among them. Driving a browser with Karate is not part of this plan, and how a load model is designed and read belongs to the Gatling roadmap; the QA Engineer role carries the rest of the loop.