- Image and information responses include a
Last-Modified
header when possible. - The health endpoint is enabled via
endpoint.health.enabled
rather thanendpoint.api.enabled
. - Added an HTTP API method to purge all infos from the derivative cache.
- Added a configuration option to automatically purge source-cached images whose format cannot be inferred.
- HttpSource supports a client HTTP proxy. (Thanks to @mightymax and @mlindeman)
- HttpSource can be configured to send a ranged GET request instead of a HEAD request, enabling it to work with pre-signed URLs that do not allow HEAD requests.
- S3Source supports multiple endpoints when using ScriptLookupStrategy.
- S3Cache uses multipart uploads, which reduces memory usage when caching derivatives larger than 5 MB.
- The delegate script pathname can be set using the
-Dcantaloupe.delegate_script
VM argument, which takes precedence over thedelegate_script.pathname
configuration key. - The delegate script's
metadata
context key contains a new field,xmp_elements
, that provides a high-level key-value view of the XMP data.
- IIIF information endpoints always return JSON in HTTP 4xx responses.
- Fixed a bug whereby the values of the
operations
andpage_count
keys in the delegate context were not set. - TurboJpegProcessor is able to generate non-JPEG derivative images, which fixes an HTTP 415 error that would occur when trying to do that.
- Fixed a crop-offset bug that could occur when using PdfBoxProcessor to generate JPEGs with libjpeg-turbo active.
- Updating libraries to fix security issues. Full details in #634
- Update of Jena to 4.8 requires RDF to have a populated rdf:about field. May impact some XMP header processing.
- Fixed an uncaught exception when returning HTTP 401 status from the
pre_authorize()
orauthorize()
delegate methods. - Updated the Logback dependency to version 1.2.8.
- Fixed a bug involving deserialization of BYTE-type EXIF fields containing numeric values.
- Fixed a bug in S3Cache that could cause source content to be purged when sharing a bucket with S3Source with the cache worker enabled.
- The
full_size
andmetadata
keys are available in the delegate context in response to information requests as well as image requests. - The application exits when it fails to bind to a port at startup.
- Improved recovery from corrupt cached infos.
- Suppressed an error-level log message from OpenJpegProcessor when reading an image without a recognized filename extension on a read-only filesystem.
- Improved the resiliency of the built-in EXIF reader.
- Fixed IIIF Image API image and information endpoints not respecting scale constraints.
- Fixed the IIIF Image API v2 and v3 image endpoints not sending a
Link
header in response to requests for cached images. - Fixed HttpSource failing to use HTTP Basic credentials when chunking is enabled.
- Fixed a potential ConcurrentModificationException when using the delegate script.
- Updated JRuby to version 9.2.17.0.
- Fixed a bug introduced in 5.0.1 that broke static asset serving.
- Fixed several issues related to serving static assets from within the JAR.
- Added several keys to the status endpoint response (
/status
). - Fixed a bug in the IIIF Image API 2 & 3 endpoints that caused the
profile.formats
key to be missing from the information responses of cached images. - Fixed the
page_count
key not being set in the delegate context when returning a cached info. - Fixed several issues related to EXIF metadata parsing.
- Fixed a floating-point rounding bug that could cause an excessively large TIFF pyramid level to be selected, resulting in unnecessary scaling.
- Fixed a bug in TurboJpegProcessor that could cause incorrect cropping.
- Fixed a bug in S3Cache that could cause hits against expired content.
- Fixed a bug that could cause S3 credentials from the configuration file to be erroneously used even when not set.
- Fixed a bug that could cause corrupt image data to be written to a derivative cache.
- Updated the PDFBox dependency to address the following security vulnerabilities: CVE-2021-27807, CVE-2021-27906.
- Added an endpoint supporting the IIIF Image API version 3.0 at
/iiif/3
. - When an image requested at
max
size is larger than the value of themax_pixels
configuration key, it is downscaled to that value instead of the request being forbidden. This behavior aligns more closely with the IIIF Image API. - The minimum and maximum size of the built-in web server's thread pool is configurable.
- Added a configuration option to control whether the health check includes dependent sources and caches.
- A
cache=recache
URL query argument is available that reprocesses and recaches the derivative image before delivering it. - The
X-IIIF-ID
reverse proxy header is no longer supported.X-Forwarded-ID
should be used instead. - The
response-content-disposition
URL query argument supports RFC 6266 Unicode filename syntax. - The
endpoint.iiif.content_disposition
configuration key has been removed. Theresponse-content-disposition
URL query argument should be used instead. - The
http.http2.enabled
andhttps.http2.enabled
configuration keys have been removed. HTTP/2 is always enabled. - The identifier path component of IIIF Image API URIs may contain a "meta-identifier," which is comprised of an identifier plus other relevant information like a page number and/or scale constraint.
- The
page
andtime
query arguments have been deprecated in favor of the meta-identifier-expressed page number explained above. - Added a
log_error_responses
configuration key that causes all errors at all endpoints to be logged, which may help to track down errors that evade logging elsewhere.
- Sources support inconsistencies between filename/identifier extensions and byte signatures.
- HttpSource uses the OkHttp HTTP client library instead of Jetty.
- HttpSource supports HTTP/2.
- S3Source uses version 2 of the AWS SDK insead of MinIO.
- ImageMagickProcessor, GraphicsMagickProcessor, and KakaduDemoProcessor have been removed.
- Added GrokProcessor. (Thanks to @boxerab)
- Java2dProcessor supports multi-page TIFFs.
- The metadata-handling system has been redesigned:
- Source image metadata is cached in derivative caches.
- XMP metadata can be copied or edited using a delegate method.
- The
processor.metadata.respect_orientation
configuration key has been removed and EXIF Orientation values are always respected.
- KakaduNativeProcessor, OpenJpegProcessor, and TurboJpegProcessor respect the
EXIF
Orientation
tag. - KakaduNativeProcessor, OpenJpegProcessor, and TurboJpegProcessor support EXIF, IPTC IIM, and XMP metadata.
- Java2dProcessor supports the XPM source format.
- Added a
processor.downscale_linear
configuration key to enable downscaling in a linear color space. - TurboJpegProcessor decompresses with "accurate" rather than "fast" DCT.
- FfmpegProcessor, KakaduNativeProcessor, OpenJpegProcessor, and PdfBoxProcessor use libjpeg-turbo to write JPEGs if it is available.
- Added the
scaled
position for image overlays. - Updated the included Kakadu library used by KakaduNativeProcessor to version 8.0.3.
- Updated Apache PDFBox to version 2.0.22.
- PdfBoxProcessor properties added to better control memory
processor.pdf.scratch_file_enabled
allows use of a scratch fileprocessor.pdf.max_memory_bytes
allows control over how much memory a PDF thread consumes (-1 means unlimited)
- S3Cache has been updated to use version 2 of the AWS SDK.
- RedisCache has been updated to use version 6.0.1 of the Lettuce client.
- The delegate method invocation cache has been removed.
- A delegate class can be written in Java. See the user manual for more information.
- Added a
pre_authorize()
delegate method that enables more efficient, "fail-fast" authorization. - Added
local_uri
andpage_count
keys to the delegate script context. - Corrected the structure of the
cookie
key in the delegate script context. See the upgrade guide for more information. - Updated JRuby to version 9.2.13.0.
- The application is now packaged as a JAR file and can no longer work in a Servlet container. (See #339 for background.)
- Java 11 is required.
- Added configuration options for logging in Logstash format. (Thanks to @cmhdave)
- The
redaction.enabled
configuration key has been removed. Theredaction()
delegate method is always called and redactions are always applied if it returns any. - The
overlay()
delegate method may return a booleanword_wrap
key value to enable automatic wrapping of overlay strings. - The
-Dcantaloupe.list_fonts
VM argument has been replaced with the-list-fonts
command-line argument.
- Fixed an issue with the 4.1.10 build whereby the logback-classic dependency had been updated to version 1.2.8 but not logback-core.
- Fixed the IIIF Image API v2 image endpoint not sending a
Link
header in response to requests for cached images. - Suppressed an error-level log message from OpenJpegProcessor when reading an image without a recognized filename extension on a read-only filesystem.
- Updated the Logback dependency to version 1.2.8.
- The identifier scale constraint suffix syntax is configurable via new
scale_constraint_suffix.format
andscale_constraint_suffix.pattern
keys. (These are a stopgap addition that are not relevant in version 5.0.) - Fixed a floating-point rounding bug that could cause an excessively large TIFF pyramid level to be selected, resulting in unnecessary scaling.
- Fixed a bug that could cause corrupt image data to be written to a derivative cache.
- Updated the PDFBox dependency to address the following security vulnerability: CVE-2021-27807, CVE-2021-27906.
- Fixed a rounding bug that could cause requests for certain scale-constrained images to return HTTP 403 status.
- Fixed a potential NullPointerException from Java2dProcessor when the
processor.metadata.respect_orientation
configuration key is set totrue
. - Improved TurboJpegProcessor's ability to partially decode corrupt source images.
- Improve the efficiency of the health check endpoint.
- Fixed a sporadic JVM crash when using KakaduNativeProcessor under load with a derivative cache enabled.
- Fixed incorrect success status in response to image requests that have failed with a VM error.
- Fixed a
NumberFormatException
caused by incorrect Java version parsing when running in a beta or early-access JVM. - Fixed an
IOException
appearing in the log at the conclusion of a successful request involving HttpSource when chunking is enabled. - Fixed a
NullPointerException
when returningnil
from theoverlay()
delegate method. (Thanks to @ccare) - Fixed an error from the embedded Jetty server when trying to use a key store with multiple certificates.
- Image information is no longer included in IIIF information responses with an HTTP 403 status.
- Fixed overridden boolean values not being recognized in inherited configuration files.
- Fixed information responses failing to respect the
page
URL query argument. - Fixed a bug that caused requests for data beyond an offset of 2^31 from S3Source and AzureStorageSource to fail when chunking is enabled.
- Updated JRuby to version 9.2.11.1, which addresses multiple CVEs.
- Updated Jackson to version 2.11.0, which addresses multiple CVEs.
- Fixed the IIIF Image API 2.x endpoint's handling of identifiers containing backslashes.
- Fixed a potential NullPointerException thrown by various AbstractIIOImageReader methods resulting from improper exception handling.
- Improved the user-friendliness of error messages resulting from illegal quality and format arguments in image requests.
- Reduced an error-level log statement resulting from an exception while streaming from a derivative cache to the response to debug-level.
- Jetty has been upgraded to version 9.4.24.v20191120, which addresses multiple CVEs.
- Fixed the health-check endpoint (at
/health
). - Fixed error responses when encountering empty request headers. (Thanks to @RRMoelker)
- Fixed incorrect URIs in
Link
header values in certain IIIF Image API v2 image endpoint responses. - The
?response-content-disposition
query argument correctly handles filenames containing spaces. - Image endpoints return HTTP 403 instead of 500 in response to AccessDeniedExceptions from the underlying Source.
- Fixed two separate bugs in KakaduNativeProcessor that both caused empty regions to appear in certain images.
- Fixed overly restrictive signature verification in KakaduDemoProcessor and OpenJpegProcessor that could cause errors with some valid images.
- Fixed TurboJpegProcessor not respecting the
processor.background_color
configuration key. - Fixed the
PurgeItemFromCache
API method not working with S3Cache and AzureStorageCache. - When using
AutomaticSelectionStrategy
, processor initialization errors are logged at debug-level rather than error-level.
- Added an undocumented
HttpSource2
source which works the same asHttpSource
, and uses the same configuration keys and delegate method, but is backed by a different HTTP client library. If you are having problems withHttpSource
, try this one instead. The current implementation ofHttpSource
will probably be replaced by this one in the next major version. - Fixed a missing
resulting_size
key in the delegate script'soptions
hash, introduced in 4.1. - Fixed a bug in KakaduNativeProcessor that could cause requests for certain scales to fail (or possibly even crash the JVM).
- Fixed the behavior of
response-content-disposition
query arguments containing afilename
argument. - Fixed failing to source-cache non-selectively-readable images when using Java2dProcessor or JaiProcessor with the source cache enabled and a source for which chunking is enabled.
- The Jetty client used by HttpSource has been downgraded to the version used in 4.1.1 due to a bug in the version used in 4.1.2.
- Fixed the behavior of the
print_stack_trace_on_error_pages
configuration key in the context of certain unrecoverable errors.
- Fixed incorrect
@id
values in information responses when a query argument is present in the URI. - Fixed a potential failure to fall back to the next candidate processor when TurboJpegProcessor is in the candidate chain but libjpeg-turbo is not available.
- Fixed a potential IIOException when reading certain JPEG images with JaiProcessor.
- Fixed newlines (
\n
) appearing literally in string overlays. - Updated the Jetty dependency to address the following security vulnerability: CVE-2019-10247.
- Updated the PDFBox dependency to address the following security vulnerability: CVE-2019-0228.
- Fixed
NoSuchMethodError
s from various processors when running in Java 8. - Fixed failure to locate the delegate script when
delegate_script.pathname
is set to a relative path outside of the current working directory. - Fixed a regression in the handling of the
X-Forwarded-Path
header. - Fixed breakage in manual processor selection from the Control Panel.
- Fixed an error from Java2dProcessor when downscaling TIFF source images to a very small size.
- Fixed improper codestream access in KakaduNativeProcessor that could cause cause degraded output for certain images.
- Improved resilience when reading certain oddly-encoded GIFs in Java2dProcessor and JaiProcessor.
- Configuration values can be obtained from the environment. See the user manual for more information.
- The
redirect()
andauthorized?()
delegate methods have been merged into a newauthorize()
method, which enables a superset of their functionality, including challenge responses and redirects to "virtual" quality-limited versions. Authorization can also be set up to align with the IIIF Authentication API 1.0's "all or nothing access" or "tiered access" schemes. - The
endpoint.public.auth.*
keys that controlled global HTTP Basic protection have also been replaced by theauthorize()
delegate method. - The maximum allowed scale can set in the configuration.
- Running tasks are displayed in the Control Panel.
- Added HTTP API endpoints for status and health checks.
- In the IIIF Image API 2.x endpoint, the
Access-Control-Allow-Origin
response header is always included in the information response regardless of whether anOrigin
header was sent in the request, in order to align more closely with the IIIF Image API. - Improved handling and documentation of the
X-Forwarded-Port
header. - The
/iiif
URI path no longer redirects to a specific Image API version, and now returns HTTP 404. - Bug fixes related to cropping-by-percent of EXIF-rotated images.
- During image requests,
authorize()
is called earlier in the setup process, for better efficiency. - Removed the Restlet dependency.
- HttpSource, S3Source, and AzureStorageSource are able to request chunks of resources on demand, which can improve performance when reading images in selectively-readable encodings like JPEG2000 and multi-resolution TIFF.
- HttpSource sends initial
HEAD
requests instead of rangedGET
requests. - The hash returned from the
httpsource_resource_info()
delegate method may include custom request headers. - HttpSource's
HttpSource.trust_invalid_certs
configuration key has been renamedHttpSource.allow_insecure
. When enabled, insecure cipher suites are accepted as well as insecure certificates. - S3Source uses a Minio client instead of the AWS Java SDK client, to work around a possible thread-starvation bug in the AWS Java SDK when S3Source and S3Cache are in use at the same time.
- Added configurable processor selection strategies, one of which uses the
existing
processor.*
configuration keys, and the other of which uses an automatic strategy that attempts to choose a "best" processor on a per- request basis. - Added TurboJpegProcessor.
- Added the
repeat
position for image overlays. - Efficiency improvements in KakaduNativeProcessor.
- KakaduNativeProcessor supports
sizeByDistortedWidthHeight
. - Java2dProcessor and PdfBoxProcessor can preserve XMP metadata across most source and output formats.
- KakaduNativeProcessor respects the
processor.metadata.preserve
configuration key for XMP metadata. - Worked around a bug in the GraphicsMagick
gm
command that causes occasional "broken pipe" errors from GraphicsMagickProcessor when reading images from a FileSource. - KakaduDemoProcessor has been deprecated, as it has been made more-or-less redundant by KakaduNativeProcessor.
- Updated the PDFBox dependency.
- Removed normalization.
- S3Cache uses last-accessed rather than last-modified times.
- Changed the S3Cache and AzureStorageCache key naming schemes to improve organization and avoid problems with edge cases.
- The change log has moved from the website to this file.
- Java 11 is supported.
- Fixed a scaling-related regression in PdfBoxProcessor.
- Streams consumed from S3Source are drained fully before closure, which makes its connection pool more efficient and fixes a warning-level error message from its internal HTTP client.
- Fixed inability to use
ScriptLookupStrategy
with AzureStorageSource. (Thanks to @marc-sensenich) - Fixed a potential math error when cropping certain images with an embedded EXIF Orientation tag.
- Improved exception handling in S3Source and S3Cache.
- Fixed inability of S3Cache and AzureStorageCache to recognize a
cache.server.derivative.ttl_seconds
key value greater than 2^31. - Worked around an issue in OpenJPEG that could cause OpenJpegProcessor to fail when reading certain images from the source cache.
- Corrected the
gray
quality togrey
in the IIIF Image API 1.x endpoint. - Updated the Tika dependency to address the following security vulnerability: CVE-2018-17197.
- Updated the Jackson dependency to address several security vulnerabilities.
- Updated the PDFBox dependency.
- The exploded application is more reliably cleaned up at shutdown when running in standalone mode.
- KakaduNativeProcessor supports JPEG2000-compatible JPX source images.
- Fixed an error from KakaduDemoProcessor in Windows when using a system locale that uses characters other than periods as decimal separators.
- Fixed incorrect redaction positioning when scaling.
- Fixed an error from processors that use a Java 2D processing pipeline when attempting to scale a source image that is natively less than three pixels in either dimension, or has been cropped to that size.
- Fixed an empty response when image overlays are enabled but the overlay pathname or URL is invalid. Instead, images are served without the overlay, and warnings logged.
- Updated Jetty to version 9.4.12, which fixes a potential connection leak in long-running servers when using HttpSource. (Thanks to @giancarlobi)
- Miscellaneous other dependency updates.
- Fixed a regression introduced in 3.4.3 that could cause URIs to appear with
the wrong scheme in information responses and
Link
headers. (Thanks to @mmatela and @jweisman) - Fixed a bug that caused the minimum tile size in IIIF Image API 2.x
information responses to be 512 regardless of the setting of
endpoint.iiif.min_tile_size
. - Fixed a connection leak in HttpSource.
- Improved the resiliency of the internal JPEG2000 parser used by KakaduDemoProcessor and OpenJpegProcessor.
- Fixed KakaduNativeProcessor's handling of
sizeByDistortedWidthHeight
requests. - Removed an error message that would appear in the log when the delegate script file was not found but the delegate script was disabled.
- The
X-Forwarded-Host
reverse proxy header may contain a port number, which will be used ifX-Forwarded-Port
is not supplied.
- A new
X-Forwarded-ID
reverse proxy header replacesX-IIIF-ID
, which has been deprecated, and will be removed in a future version. - Error response representations are in
text/plain
format in the absence of a client preference. - When
endpoint.iiif.content_disposition
is set toinline
, theContent-Disposition
header value includes a filename-safe version of the image identifier.
- Resolvers have been renamed sources to better reflect their core function.
- AmazonS3Resolver has been renamed S3Source, and it now supports non-AWS endpoints.
- S3Source supports a configurable path prefix and suffix when using
BasicLookupStrategy
. - AzureStorageSource supports shared access signature (SAS) URIs. (Thanks to @ddisciascio)
- Improved the efficiency of source cache content access when
cache.server.resolve_first = false
. - All sources fall back to checking the magic bytes in image data if they cannot infer a format any other way.
- KakaduProcessor has been renamed KakaduDemoProcessor to reflect the fact that
it interfaces with the Kakadu demo tool,
kdu_expand
, and not the native library. - Added KakaduNativeProcessor, as well as compiled binaries for Linux, macOS, and Windows, which are free to use with it for non-commercial purposes.
- Added
DownloadStrategy
as another optional retrieval strategy for stream-based processors, and as an optional fallback strategy. This enables all processors to work with all sources with no extra configuration. - JPEG2000-compatible JPX images are readable by all processors that support JPEG2000.
- JaiProcessor has been deprecated. See the user manual for more information.
- KakaduDemoProcessor and OpenJpegProcessor work in Windows.
- KakaduDemoProcessor and OpenJpegProcessor read image metadata using custom
code, which is more efficient than using the respective
kdu_jp2info
andopj_dump
tools, and enables them to read the number of decomposition levels, which improves reliability when decoding images with less-common level counts. - Image I/O plugins can be selected on a per-format basis.
- Improved the efficiency of the Java 2D resample filters.
- When using the Java 2D pipeline to downscale to less than three pixels on a side, an empty image with the correct dimensions is returned, rather than a downscaled 3×3 image, which is the smallest size that the Java 2D resample filters support.
- Java2dProcessor supports animated GIFs.
- Java2dProcessor supports CMYK & YCCK JPEGs.
- Worked around some behavior of the
opj_decompress
tool that could cause OpenJpegProcessor to break. (Thanks to @adam-vessey) - Improved PdfBoxProcessor's validation of requests for multi-page PDFs whose pages have inconsistent dimensions.
- The
processor.limit_to_8_bits
configuration option is no longer available. All output is limited to 8 bits. - Disabled support for GIF source images in JaiProcessor, as it was buggy and other processors can handle GIF better.
- AmazonS3Cache has been renamed S3Cache, and it now supports non-AWS endpoints.
- The time-to-live of the source and derivative caches is independently configurable.
- The cache worker dumps the contents of HeapCache during its shifts, if both are enabled and if HeapCache persistence is also enabled.
- The cache worker's work interval specifies the amount of time between the end of one shift and the beginning of the next, rather between the start of one shift and the next.
- The deprecated
DELETE /cache/:identifier
HTTP API method has been removed.
- The delegate script has been redesigned as a class that is instantiated per-request. See the "Delegate Script" section of the user manual for detailed information.
- Java 10 (18.3) is supported.
- Updated JRuby to 9.1.17.0.
- Fixed a regression introduced in 3.4.3 that could cause URIs to appear with
the wrong scheme in information responses and
Link
headers. (Thanks to @mmatela and @jweisman) - The
X-Forwarded-Host
reverse proxy header may contain a port, which will be used ifX-Forwarded-Port
is not supplied.
- Updated the Jackson dependency to address the following security vulnerabilities: CVE-2018-7489, CVE-2018-5968, CVE-2017-15095.
- AmazonS3Resolver and AmazonS3Cache can obtain credentials from Amazon container environment variables or EC2 Instance Metadata Service.
- Arbitrary
X-Forwarded
headers are respected withoutX-Forwarded-Host
needing to be set. (Thanks to @jweisman) - Fixed a potential resource leak involving stream-based resolvers.
- Fixed HttpResolver's handling of semicolon-delimited
Content-Type
header values. - Fixed palette corruption when cropping source images with indexed color with Java2dProcessor.
- Fixed a
NullPointerException
when using the info cache without a derivative cache. - Other minor fixes.
- Fixed a leak when using HttpResolver in a Servlet container that resulted in an inability to shut down properly. (Thanks to @punkfairy)
- Fixed aspect-fit scale-to-height not working properly in ImageMagickProcessor and GraphicsMagickProcessor.
- Fixed incorrect URIs in responses when using slash substitution or URI-encoded identifiers.
- Efficiency improvements in AmazonS3Resolver and AzureStorageResolver.
OutOfMemoryError
s during image responses are logged at error level rather than info level.- Improved the reliability of Image I/O plugin selection.
- Updated the PDFBox dependency to version 2.0.8.
- Improved efficiency when handling a request for a bitonal or gray image when the source image is already bitonal or gray. (Thanks to @adam-vessey)
- Fixed upscaling by percent in KakaduProcessor and OpenJpegProcessor.
- Efficiency improvements in PdfBoxProcessor.
- Improved the reliability of source cache downloading and content acquisition.
- Miscellaneous reliability improvements in AmazonS3Resolver and AmazonS3Cache.
- Server misconfigurations cause HTTP 500 (Internal Server Error) responses rather than HTTP 400 (Bad Request).
- HTTP/2 can be enabled in standalone mode. (H2 requires Java 9, but H2C works in Java 8.)
- The accept queue limit of the standalone HTTP & HTTPS servers is configurable.
- The minimum size in IIIF Image API 2.x
info.json
responses is configurable. (Thanks to @cbeer) - The maximum size in IIIF Image API 2.x
info.json
responses is the full image size, rather than half size. (Thanks to @cbeer) - The IIIF Image API 1.x endpoint returns a JPEG image, rather than GIF, when
no client preference is expressed in the URI or
Accept
header. - Certain derivative cache errors can be recovered from without causing an error response.
- The
sizes
key in Image API 2.x information responses respects themax_pixels
configuration option. - Image and information endpoints return
HTTP 501 Not Implemented
, rather than500 Internal Server Error
, when the selected processor does not support the requested source format. - Image requests may include a
?response-content-disposition
query argument to suggest a responseContent-Disposition
header value. - Added several HTTP API methods for cache management.
- Changes to HTTP Basic authorization settings in all endpoints take effect without restarting.
- All endpoints support the HTTP
OPTIONS
method. - Status items in the Control Panel update automatically via AJAX.
- The username used to access the Control Panel is configurable.
- Miscellaneous efficiency improvements across all resolvers.
- The
HttpResolver::get_url()
delegate method may return a hash containing authentication credentials. - Added an option to HttpResolver to trust all certificates.
- The request timeout of HttpResolver is configurable.
- AmazonS3Resolver and AzureStorageResolver attempt to infer a source format from the object/blob key if not able to do so from the identifier or from object/blob metadata.
- The maxmimum number of AWS connections used by AmazonS3Resolver is configurable.
- AmazonS3Resolver falls back to obtaining credentials from various other sources (see the user manual) when they are not set in the configuration file. (Thanks to @jweisman)
- Resolver delegates are passed an additional context parameter. (Thanks to @jonathangreen)
- Added an option that enables Java2dProcessor, JaiProcessor, KakaduProcessor, and PdfBoxProcessor to preserve the sample depth of >8-bits-per-sample source images.
- GraphicsMagickProcessor and ImageMagickProcessor respect the
processor.dpi
configuration key. - GraphicsMagickProcessor supports the GIF output format.
- ImageMagickProcessor supports image overlays.
- FfmpegProcessor supports FLV videos.
- The
processor.background_color
configuration key affects alpha blending as well as non-90°-multiple rotations. (Thanks to @Algae666) - ImageMagickProcessor's support for ImageMagick versions prior to 7 is deprecated.
- OpenJpegProcessor's support for OpenJPEG versions prior to 2.2.0 is deprecated.
- Improved the predictability of Image I/O plugin selection.
- Initialization warnings (related to missing dependencies, for example) are displayed in the Control Panel.
- Image info is written to caches asynchronously to enable faster responses.
- Added an optional in-memory image info cache.
- Added RedisCache.
- Added HeapCache.
- AmazonS3Cache uploads images to S3 asynchronously, enabling faster responses.
- The maxmimum number of AWS connections used by AmazonS3Cache is configurable.
- AmazonS3Cache falls back to obtaining credentials from various other sources (see the user manual) when they are not set in the config file. (Thanks to @jweisman)
- On macOS, FilesystemCache uses file last-modified times rather than last-accessed times as the latter are not reliable.
- Configuration files support inheritance via an
extends
key. - The temporary directory is configurable.
- Added an optional error log.
- A logger is accessible from the delegate script.
- Fixed FilesystemCache failing to purge invalid source images. (Thanks to @jhosteny)
- Fixed the
cache.server.resolve_first
configuration key always being effectivelytrue
in the IIIF Image API 2.x endpoint. - Fixed AmazonS3Cache sometimes purging fewer objects than it should.
- Fixed failure to auto-purge incompletely-written images from derivative
caches other than FilesystemCache in response to a VM error (such as
OutOfMemoryError
). - HttpResolver makes fewer HTTP requests.
- Fixed an
AbstractMethodError
in HTML views, caused by a regression in 3.3.3. - Fixed the
X-IIIF-ID
header not being respected when redirecting from/:identifier
to/:identifier/info.json
in both IIIF Image API endpoint versions. (Thanks to @jonathangreen)
- Added Java 9 compatibility.
- Upgraded dependencies to fix the following vulnerabilities: CVE-2015-6420. (Thanks to @kinow)
- Fixed incorrect computation of post-rotation bounding boxes, manifesting as
bogus values passed to the
resulting_size
delegate method arguments. (Thanks to @SaschaAdler) - Fixed a warning from FilesystemCache when multiple processes tried to create the same directory at the same time. (Thanks to @cbeer)
- Fixed potentially incorrect tile sizes in information responses when using OpenJpegProcessor.
- Fixed a bug in FfmpegProcessor that caused video info parsing to fail for some videos.
- Added
Origin
to allVary
response headers, which fixes a potential CORS-related caching issue. - Fixed an issue where certain Image I/O plugins could fail to register when multiple applications that use Image I/O are running in the same container. (Thanks to @mmatela)
- Fixed a resource leak when using image overlays.
- Improvements in the testing and build process. (Thanks to @kinow)
- Fixed an inefficiency that caused image metadata to be read multiple times in a single image response.
- Fixed endpoint authentication being enabled for certain non-endpoint resources.
- Impoved compatibility between OpenJpegProcessor and OpenJPEG when using OpenJPEG version 2.2.0 or later. (Thanks to @RussellMcOrmond)
- Fixed inability to stream source images when using AmazonS3Resolver.
- Fixed a potential HTTP 500 error when multiple processes try to write the same image metadata concurrently using FilesystemCache.
- Corrected the documentation for
processor.upscale_filter
andprocessor.downscale_filter
. - GraphicsMagickProcessor and ImageMagickProcessor support a
?page
query argument with PDFs.
- Fixed caches failing to take URL query arguments into account.
- Fixed a potentially incorrect
Cache-Control
header being sent in HTTP ≥ 400 responses. - Fixed parsing of comma-separated
X-Forwarded-*
header values. - Improved the reliability of the value of the
client_ip
argument to theauthorized?()
delegate method when behind a reverse proxy. - Fixed a connection leak in AmazonS3Resolver.
- Fixed potential malformed concurrent image responses when using a processor other than ImageMagickProcessor or GraphicsMagickProcessor.
- Fixed a potential NullPointerException (HTTP 500) involving concurrent requests when using the delegate script.
- Fixed processors and caches not respecting the value of
processor.normalize
. - Prevented the configuration file and/or delegate script from being reloaded twice on some platforms.
- Fixed PDF source format support in GraphicsMagickProcessor and ImageMagickProcessor.
- Upgraded JRuby to 9.1.10.0.
- Auto-computed tile sizes of source images with an odd number of pixels in a dimension in information responses are rounded up, to prevent clients from requesting extremely narrow edge tiles.
- Operations that would have no effect in their context—for example, a crop to 0,0/300×200, followed by a scale-to-fit to 300 pixels wide—are skipped.
- Improved request validation.
- Added a RESTful API endpoint to retrieve and update the configuration.
- Most processor-specific configuration settings now apply across all processors, making them integrate better into the caching architecture.
- Added an auto-normalize option to all image processors.
- Added an option to all processors to output progressive JPEGs.
- Added experimental support for DICOM images to GraphicsMagickProcessor, ImageMagickProcessor, Java2dProcessor, and JaiProcessor. (This is a very complex format and not all variants are working.)
- FfmpegProcessor uses a Java 2D processing pipeline, giving it more capabilities.
- Java2dProcessor and JaiProcessor support 16-bit LZW- and ZIP-compressed TIFFs.
- GraphicsMagickProcessor and ImageMagickProcessor support selectable resample filters.
- GraphicsMagickProcessor and ImageMagickProcessor respect the EXIF Orientation tag.
- ImageMagickProcessor supports ImageMagick 7.
- PdfBoxProcessor supports JBIG2-encoded embedded images.
- PdfBoxProcessor returns HTTP 400 when given an illegal page number.
- FfmpegProcessor returns HTTP 400 when given an illegal time.
- The
-Dcantaloupe.cache.purge
VM option accepts an identifier argument to purge all content related to an individual identifier. - Added a RESTful API endpoint to purge individual images from the cache.
- A
cache=false
URL query option can be used in all endpoints to bypass the derivative cache and disable theCache-Control
header per-request. - FilesystemCache supports identifiers longer than the filesystem's filename length limit.
- Watermarks have been renamed overlays.
- Added textual overlays.
- Overlay images can be located on the filesystem or an HTTP(S) server.
- Overlay images are cached in memory for better efficiency.
- Added an optional delegate method invocation cache.
- Added the ability to redirect to another URL via the
authorized?()
delegate method. - The delegate script is reloaded on change rather than per-request.
- Fixed all processors except JaiProcessor treating scale-to-dimensions
requests (
w,h
) as scale-to-fit (!w,h
). - Fixed a bug affecting upscaled crops when using pyramidal images with Java2dProcessor or JaiProcessor.
- Worked around a bug in JAI involving downscaling of compressed TIFF tiles, affecting JaiProcessor.
- Fixed inability to upscale in JaiProcessor, omitted in error from 3.2.2.
- Fixed incorrect scale size computation using flooring instead of rounding.
- Fixed a bug in FilesystemCache that could cause an HTTP 500 error when
accessing the same uncached
info.json
resource concurrently. - FilesystemCache cleans up zero-byte files created in certain situations, typically involving failure in the underlying storage.
- Fixed a bug that could cause incompletely-written cached images (with any cache) to not get cleaned up properly.
- Fixed inability to upscale by pixels.
- Fixed an error in response to requests for images less than 3 pixels on a side when sharpening is enabled, in all processors using the Java 2D pipeline.
- Fixed an error in JaiProcessor when attempting to scale an image less than 3 pixels on the smallest side.
- Fixed a bug in FilesystemCache that caused zero-byte files to be written and an HTTP 500 response when handling concurrent identical requests.
- Error responses include a more appropriate
Cache-Control
header rather than the one set in the configuration. - Trailing slashes in
X-Forwarded-Path
header values are trimmed off, preventing double-slashes from appearing in generated URLs.
- Fixed a bug that caused the resolver selected by the
get_resolver()
delegate method to not be initielized properly. - Fixed a path-handling bug affecting ImageMagickProcessor and GraphicsMagickProcessor primarily in Windows.
- Fixed incorrect image sizes when using FfmpegProcessor with percentage scales above 100%.
- Improved support for 16-bit source images in Java2dProcessor, JaiProcessor, GraphicsMagickProcessor, and ImageMagickProcessor.
- Java 8 is now required.
- Significant efficiency improvements in the Java 2D processing pipeline affecting Java2dProcessor, JaiProcessor, and KakaduProcessor.
- All processors that use the Java 2D pipeline (Java2dProcessor, KakaduProcessor, OpenJpegProcessor, PdfBoxProcessor) support selectable upscaling and downscaling filters.
- Improved the scale quality of JaiProcessor.
- All processors offer configurable sharpening levels.
- Applicable processors respect source ICC color profiles.
- Added the option to copy EXIF, IPTC IIM, XMP, and/or native TIFF and PNG metadata from source images to derivative images of the same format, for some processors.
- Java2dProcessor and JaiProcessor respect the value of an embedded EXIF Orientation tag.
- Increased the default Java2dProcessor/JaiProcessor JPEG quality to 0.8 (80).
- Added support for a
X-IIIF-ID
request header, in order to return correct image identifiers in information responses whose requests have passed through a reverse proxy using a different identifier. - Removed the JAI post-processor option for KakaduProcessor and OpenJpegProcessor.
- Zero-byte files are deleted when purging or cleaning FilesystemCache.
- KakaduProcessor and OpenJpegProcessor return a less inscrutable error message when they are unable to read a source file.
- HTTP 415 is returned instead of HTTP 400 when an output format is requested that is valid but unsupported by the current processor.
- Colorized the console application log.
- Console application log messages with a severity of
WARN
and above are routed to standard error. - Upgraded a number of libraries.
- Fixed a resource leak that would cause landing pages, error pages, and the Control Panel to become inaccessible after a long time running.
- Fixed a bug where the watermark wouldn't be added if redactions were enabled
and set to use
DelegateScriptStrategy
but the delegate script was disabled. - Fixed tile sizes in the IIIF Image API 2.x information response when using KakaduProcessor.
- HTTP 400 is returned in response to a request for an image that would be
zero pixels on a side, even when the zero dimension is not explicit in the
request. (For example, a size of
4,
for an image that is at least five times wider than it is tall.) (Previously, it was only returned when zero pixels was explicitly requested, and the image returned for implicit zero sizes was incorrect.)
- Added the
http.host
andhttps.host
standalone mode configuration options, defaulting to listen on0.0.0.0
(all interfaces).
- Added support for version 2.1 of the IIIF Image API.
- Added support for the
square
region parameter to all processors. - Added the
max
size parameter. - Removed the information request
Link
header for the JSON-LD context. - Added the
sizeByConfinedWh
andsizeByDistortedWh
feature names, supported by all processors. - Added the
maxArea
property to the profile description, based on the value of themax_pixels
configuration key.
- Added support for the
- Replaced the
get_iiif2_service
delegate method withextra_iiif2_information_response_keys
.
- Fixed a broken build in standalone mode.
- Added a Control Panel for web-based configuration changes.
- Changed the distribution archive to a custom WAR file that can be run standalone (as before) or in a Servlet container.
- Added AmazonS3Cache.
- Added AzureStorageCache.
- Added redaction.
- Added the ability to restrict available sizes to those specified in the
info.json
response in the IIIF Image API 2.0 endpoint. - Added an optional source image cache, which can improve the performance of StreamProcessors, as well as enable FileProcessors to work with StreamResolvers.
- Moved JdbcResolver delegate methods to the delegate script.
- Moved resolver-related delegate methods to submodules to improve organization.
- Changes to the logging configuration are applied immediately without restarting.
- Requests for more than
max_pixels
receive an HTTP 403 response rather than HTTP 500. - Upgraded the Amazon S3 client to version 1.10.56.
- Upgraded the Restlet framework to version 2.3.7.