- Adjust maximum partitions in serializer. A bug was found: text that was large but but still reasonably sized (it more than ten thousand lines) was causing a failure in templating. The serializer utility allowed a maximum of ten thousand "partitions" (i.e. lines) before throwing an exception. While it does make sense to set a maximum on any loop, ten thousand was too low. This value was raised to ten million, based on the concept that sending an HTML template with a million lines would be absurdly high, and then putting a safety factor multiple on top of that.
- Extra testing
- New ability to inject code before and after regular endpoint processing. See registerPreHandler and registerLastMinuteHandler in WebFramework.
- Extra testing across the board.
- Improve documentation.
- Remove dead code in ExtendedExecutor.
- Improve performance test for templating. Uses parallel processing to show more realistic speed.
- Include indent when rendering template. Often, a user will render a value with newlines in a template. This adjustment tracks the indent of the key in the template and then applies that to subsequent lines of the key values.
As an example, with a template like foo bar {{ color }}:
before:
foo bar red
blue
orange
after:
foo bar red
blue
orange
- bug fix: multipart data partitioning. The multipart-form decoding was not properly trimming the last two bytes - carriage return and line feed. That is to say, data being received by multipart-form had two bytes appended at the end.
- Adjust tests to work on Mac
- Include "throws Exception" on ThrowingRunnable. This obviates handling checked
exceptions when adding endpoints - the thrown exception will get caught and
logged at
throwingRunnableWrapper.
-
Deeper testing - nearly 100% test coverage
-
Breaking changes (see migration guide):
- Spelling of constant values
- Name of the HTTP codes
- Removal of the
updatemethod inDb - FunctionalTesting now requires the hostname and port
FunctionalTesting.sendnow requires a byte array body parameter- Methods in
FileUtilsnow don't throwIOException TheBrig.sendToJailnow returns true if succeededTheBrig.getInmatesnow returns a list ofInmateinstead ofList<Map.Entry<String, Long>>FullSystem.closebecomesFullSystem.shutdown
-
Refactorings / documentation improved all over
-
New helper methods in TestLogger for searching logs
-
In Logger, if the
ActionQueueis stopped or null, will fall back to useSystem.out.printf -
In TestLogger, new method
doesMessageExistwhich handles the common usage better, that is:assertTrue(logger.findFirstMessageThatContains("foo foo did a foo").length() > 0);becomes
assertTrue(logger.doesMessageExist("foo foo did a foo")); -
Truncate timestamp in logging to microseconds
-
Modify the
ThrowingRunnableand other functional interfaces to handle exceptions better -
Raised the bar for testing throughout the system. Methods were adjusted to expose code for testing. Linting tools recommended naming modifications.
-
Documentation was written to help users through migration - see docs/migration_to_v3.md.
-
The
updateandwritemethods in the database shared so much functionality, it made sense to combine them. There are now only two methods in Db handling data modification. If creating new data, set the index to 0 on the object extending DbData.This code creates a new item in the database:
db.write(new Foo(0, 2, "a"));This code updates an item:
db.write(new Foo(1, 2, "a"));Adding a new item with a positive index will fail - the database has to generate the index for you. The returned value from
writewill provide the new index. -
ThrowingRunnable and other functional interfaces now properly handle exceptions. It is no longer necessary to try-catch a checked exception when using one - anything thrown will be logged. See ThrowingRunnable.
-
FullSystem.closerenamed toFullSystem.shutdown -
Headers.makenow just requires one parameter,Context -
RequestLine.EMPTYrenamed toRequestLine.empty -
Bug fixes
- Improve testing, refactoring
- Deeper testing on InputStreamUtils
- Rudimentary testing for FullSystem
- Improved configuration for mutation tests
- Adjust port in web tests to avoid conflicts with other tests
- Documentation improvements
- Edge case improvements. Refactorings.
- If the client sends us a message body that has a different length than the content-length header, we will handle it by just logging to debug level.
- If we have determined a client is trying to attack us again while already in the brig, we will update their duration.
- If a ForbiddenUseException bubbles up to the top, we'll add that client to the brig.
- Refactoring Body - does not use a Context object, so remove it.
- Add GZIP compression. This does not require any modifications by users - but now the Minum web server will reply with GZIP compression if the client browser sends a request that includes an accept-encoding header with a value that includes "gzip". This will only apply to text data.
- Adjust documentation for ActionQueue.enqueue. The documentation was incorrectly suggesting it was necessary to return null.
- Add failure message option for all assertions. This way it is possible to include better help when a test fails, across all assertion types.
- Added a pre-filter to avoid nulls in SearchUtils.findExactlyOne, which helps avoid some null pointer exceptions.
- Remove ParsingException from being thrown in the BodyProcessor. If any issues take place with parsing the body, it will be logged as a debug issue and the body will continue with processing, just without having determined any key-value pairs.
- cannot rely on synchronized code with virtual threads. The hope was that using the synchronizedMap method would provide thread-safe access to this data. That has not proven out, and the guide to virtual threads in Java specifically warns against use of the synchronized keyword.
- It was found that when a client requested a directory, the system would throw an IOException and log an async_error. This fixes that, so that it is merely "file not found".
- New ability to inject custom properties into Constants, which will provide some better flexibility when testing. For example, you may want to run concurrent tests with different server ports, or different database folders.
- TestFramework.buildTestingContext now allows injecting a properties file to use when creating the Constants object.
- Make Constants.getConfiguredProperties public to enable tests to more easily use all the default properties, with programmatically-defined customizations
- Make code for converting comma-delimited strings to array more robust. Now, extra spaces are much less likely to confuse it.
- Provide ability to get all the keys returned in the key-value pairs of a Response body: body.getKeys(). Useful for some situations where you are dealing with dynamic fields.
- Documentation improvements.
- Increased default for maximum tokens allowed to 1000. This basically corresponds to how many fields you can have on a page. It is customizable in the configuration at MAX_TOKENIZER_PARTITIONS
- Update version numbers for some Maven plugins, allowing nicer reports, etc.
- Add Template HTML element
- Fix to bug in templating code. It was not handling an unmatched double-closing-bracket properly. Also adjusting so if there is a double-opening-bracket without a closing set, it will throw a new custom exception.
- Fix to bug in output encoding for HTML attributes, in StringUtils.encodeAttr(). There were a couple extra characters that needed encoding.
- Milder complaint if user lacks minum.config file. Before, if the user did not have a minum.config in the directory where they started Minum, it would halt with an error message showing text for a configuration. Now, the code will continue on with a warning about the missing configuration and instead use reasonable defaults.
- fixing a bug in the configuration settings for extra mime types ("EXTRA_MIME_MAPPINGS"), where a lack of value for that property would cause the system to fail on startup.
- Updated the default value for MAX_READ_LINE_SIZE_BYTES from 500 to 1024, along with an adjustment to the error message shown in the logs when the max was encountered. To provide context: this property exists to set a reasonable limit to what a maximum header size could be, to prevent certain security attacks, or to properly handle broken user agents. It was noticed that on localhost, testing against multiple servers could cause so many cookies to exist that it would exceed the 500 byte limit. Now, it is clearer when this limit has been encountered, and the value is adjustable in the configuration - see MAX_READ_LINE_SIZE_BYTES in minum.config.
- Extra documentation in methods, and added a tutorial "getting started"
- Added a new constant to control the number of elements in the file LRU cache, called MAX_ELEMENTS_LRU_CACHE_STATIC_FILES
- Renamings to better align with HTTP specification
- Better correctness of database code, prevent some race conditions through locking
- Linter suggestions
- Improved documentation
- Improved templating - less strict with whitespace around keys
- Adjust default time for use in TheBrig
- Use webapp for static and template files
- Remove StaticFileCache - see 25763cfe
- Incorporate Maven
- Improve testing - using JUnit rather than custom test framework
- Various bug fixes, refactoring, and documentation
- Adjust to allow testing virtual threads on Windows - needed sleeps after closing sockets
- Rename app.config to minum.config
- Make most classes "final" to make it clear they are not expected to be extended
- Better form/multipart handling - now provides headers per partition
- Adjust files to match typical Java/Maven patterns
- Beta release