# Movie Script parser Implemented with spring boot spring integrations and engineered toward a lambda architecture Project is maven pom based, so to test just mvn test and to run just mvn spring-boot:run The processing is divided into multiple modules and layers, making it easy to develop and test in a team greater than just myself. That way a TDD approach is feasible. I choose to use separated spring context xml configuration to make it clear the layers and that way it is easier to control and fine tune. The processing begins splitting the file into chunks of processing per movie setting: ## Setting Splitter (SettingSplitter.java) Do line breaking processing putting line into a List and extract setting name splitting file into a collection of settings for further processing *SettingSplitterTest test for a simple scenario where Setting class is filled with expected content ###Configuration: spring-script-context.xml ## Script Parser (ScriptParser.java) Do per Movie Setting parsing for Movie Characters and Dialogs, also do word splitting and consolidate per Movie Setting word counts. Use a Map index for fast retrieval of Movie Characters and consolidation. After Setting Splitter each Setting is dispatched to a task executor in a Thread pool configurable in spring-global-context making it a multi core processing unit. *ScriptParserTest test for a simple successful scenario *TrimPontuationTest test for getting rid of pontuation in words. *BlankPaddedTest for testing some scenarios on blank spaced character names and dialogs for easy parsing Helper.java : Collection of lambda functions for String processing ### Configuration: spring-script-context.xml, spring-global-context.xml ## Persist Setting (PersistSetting.java) Save results into relational database and when there is a result update with sum per Setting and per Character. As the problem is a process once and read always I have designed a solution to read consume low cpu and scales into high throughput reading. It could be faster with a document oriented db like mongodb but for now we are using in memory relational db. I have put support for optimist lock using a version column in word table, but time was not enough to further optimize processing in parallel so persisting is configured in a queue with a single thread for now. Persisting is taking the bulk of latency here so I imagine to use per node workare for high throughput processing but time was not enough and dropped the idea. *PersistSettingTest test for a number of possible paths *IntegrationTest1 test for a number of possible paths including prior processing units ### Configuration: spring-setting-context.xml ## Get Settings (GetSettings.java) Do consult on database for all Settings and further serialize into json for delivery It is using spring data jta repositories with custom queries and pagination to limit the 10 words and order desc use case *GetSettingsTest is testing for simple successful scenario, it uses some of the initial services to make developing test faster so, no mocking, it is a partial integration test ## Get Setting by ID (GetSetting.java) Almost mirror of GetSettings but for just one Setting Id *GetSettingTest tests for simple successful scenario ## Get Characters (GetCharacters.java) Do consult on database for Characters and further serialize into json for delivering *GetCharactersTest tests for simple successful scenario ## Get Character (GetCharacter.java) Almost mirror of GetCharacter but for just one Id ### Configuration: spring-getter-context.xml ## Inbound Endpoints (GetCharacterInbound.java, GetCharactersInbound.java, GetSettingInbound.java, GetSettingsInbound.java, PostScriptInbound.java) Do http use cases and call corresponding services. At the time of this writing a found a glitch into spring-boot-RC1 that I choose to use, but I have no time to workaround. The Get /characters endpoint is always failing with a nasty bug despite being on eye identical code to other endpoints :( Observations: Some tests are time and machine performance related and could fail in a slower machine, usually running each test on isolation always pass.