Skip to content

Latest commit

 

History

History
361 lines (265 loc) · 13.7 KB

NEWS.md

File metadata and controls

361 lines (265 loc) · 13.7 KB
Raw

New in version 1.52

  • Updated code to use C++17 features and new wreport/dballe
  • Fixed build/test errors (#337, #339)
  • Removed support for CentOS7 in specfile

New in version 1.51

  • Fixed scanning GRIB1 files with missing level parts (#326)
  • Removed docker references, updated build instructions (#329)

New in version 1.50

  • Implemented date/time validation in matcher parser (#324)
  • Add a formatter for annotating cosmo model local variables (#300)
  • Updated grib2 default table version to 15
  • Updated match-alias.simc.conf file with ICON grib2 products

New in version 1.49

  • arki-dump --info now documents the alias file configuration (#323)

New in version 1.48

  • Removed embedded version of sqlite (#194)
  • Removed embedded version of liblzo (#195)
  • Fixed mentions of .archive in documentation (#288)
  • Add missing files (related to JPEG scanning) in installation scripts (Meson and autools)
  • JPEG scanner doesn't raise an exception if datetime tag is missing
  • Do not install vm2/source.lua (#310)
  • Documented match-alias.conf syntax (#318)
  • Improved dealing with missing values in GRIB2 metadata (#317)

New in version 1.47

  • Fix installation scripts (Meson and autotools) #304

New in version 1.46

  • Do not accidentally create missing segments when trying to remove data from them (#296)
  • Implemented a timeout waiting for the filter child process to exit (#301)
  • Removed CentOS 7 support
  • Implemented arki-maint unarchive (#297)
  • Stop at the first grib table file found (partially fixed #300)

New in version 1.45

  • Ported to GEOS C API (#291) (#294)
  • Compatibility with newer werkzeug (#293)
  • Compatibility with gcc-c++ 12 (#292)

New in version 1.44

  • Only allow GET in dataset-specific /query if style=postprocess (#289)

New in version 1.43

  • Only allow GET in /query if style=postprocess (#289)

New in version 1.42

  • Restored HTTP GET for /query (#289)

New in version 1.41

  • Faster deletion of data from iseg datasets (#286)

New in version 1.40

  • Added documentation of metadata types (#280)
  • Fixed arki-web-proxy (#281)
  • Added initial support for JPEG files (#277)
  • Added default timeout of 15m for ARKI_IO_TIMEOUT (#270)
  • Removed vestigial link to non-existing "Perform a query" form in arki-server (#144)
  • arki-server now gives HTTP error 405 "Method not allowed" when appropriate (#144)
  • Improved rounding in ODIM scanning (#283)

New in version 1.39

  • In arki-check output, show the amount of data that would be freed by a repack (#187)
  • Fixed inconsistent handling of empty directories as dir segments (#279)

New in version 1.38

  • Dismissed type=ondisk2 dataset support (#275)
  • Implemented alias management in arki-web-proxy (#272)
  • When a config file is found inside a directory, as if it were a dataset, but has type=remote, its path value does not get overridden with the full path to the directory (#274)

New in version 1.37

  • Log Python scanner exceptions as warnings, since processing proceeds with a base set of metadata (#273)

New in version 1.36

  • arki-scan exit status now discerns not imported condition (#129)
  • arki-mergeconf now supports explicited config files as a source (#265)
  • Differentiated satellite GRIB2 products (#268)
  • Fixed detecting end of read/write flush loop (#269)

New in version 1.35

  • Added support for libnetcdf.so.15
  • Minor fixes in dependency checking and tests (#266, #267)

New in version 1.34

  • Fixed satellite grib2 import (#264)
  • Implemented server-side timeout for arki-queries (#252)

New in version 1.33

  • Fixed arki-bufr-prepare --usn (#260)
  • Improved doc (#261, #235)

New in version 1.32

  • Added missing nc.py install (#257)

New in version 1.31

  • Fixed a serious issue in encoding/decoding GRIB1 timeranges (#256)
  • Added eatmydata dataset config option documentation (#233)

New in version 1.30

  • Added initial NetCDF support (#40)
  • Metadata refactoring to address memory issues (#242, #245)
  • Implemented eatmydata: yes in dataset conf (#233)
  • Documented supported dataset configuration options (#143)
  • VM2 data: implemented a normalisation function, fixed smallfile issue (#237)
  • Fixed segfault in python test suite (#254)
  • Implemented arki.Matcher.merge(matcher) (#255)
  • Implemented arki.Matcher.update(matcher)
  • Refactored ValueBag/Values with a simple structure (#248)
  • Break circular dependency by decoupling Session and Pool (#250)
  • Fixed segfault when query data is appended to a file from a iseq dataset (#244)

New in version 1.29

  • Fixed error reading offline archives (#232)
  • Allow to create an arkimet session with force_dir_segments=True to always store each data in a separate file, to help arkimaps prototyping.
  • New eatmydata = yes dataset configuration (disabled by default) to turn off multi-process dataset access and consistency guarantees on the datasets in case of system crashes, in favour of speed. This is useful when creating, for example, a temporary work dataset in a temporary directory (#233)

New in version 1.28

  • Added format_metadata.rst to Python HOWTOs (#230)
  • Python: added arkimet.Summary.read_binary (#230)
  • Python: added arkimet.Summary.read_yaml (#230)
  • Python: added arkimet.Summary.read_json (#230)
  • Python: added annotate: bool = False to arkimet.Summary.write (#230)
  • Python: added annotate: bool = False to arkimet.Summary.write_short (#230)
  • Python: implemented arkimet.Metadata.write for yaml and json formats (#230)
  • Python: added annotate: bool = False to arkimet.Metadata.write (#230)
  • Python: added arkimet.Metadata.read_yaml (#230)
  • Python: added arkimet.Metadata.read_json (#230)

New in version 1.27

  • reftime match expressions using repeating intervals (like reftime:=yesterday %3h) now have a new @hh:mm:ss syntax to explicitly reference the starting point of the repetitions. For example, arki-dump --query 'reftime:=2020-01-01 20:30%8h' yields reftime:>=2020-01-01 20:30:00,<2020-01-01 20:31:00,@04:30:00%8h, meaning a match in that interval and at 20:30 every 8 hours. (side effect of fixing #206)
  • Added Python bindings for arki.dataset.Session (side effect of fixing #206)
  • Added Python bindings for arkimet.dataset.Dataset (side effect of fixing #206)
  • Python code should now instantiate datasets and matchers using Session; the old way still works, and raises DeprecationWarnings
  • Python function arkimet.make_qmacro_dataset() deprecated in favour of arkimet.dataset.Session.querymacro()
  • Python function arkimet.make_merged_dataset() deprecated in favour of arkimet.Session.merged()
  • Python function arkimet.get_alias_database() deprecated in favour of arkimet.dataset.Session.get_alias_database()
  • Python function arkimet.expand_query() deprecated in favour of arki.dataset.Session().expand_query()
  • Correctly query multiple remote servers with different but compatible alias configurations, in merged and querymacro datasets (#206)

New in version 1.26

  • Fixed use of geos compile-time version detection (#225)
  • Removed make install depending on make check (#218)
  • Fix ODIM scanner for volumes with missing 'radhoriz' parameter (#227)

New in version 1.25

  • Fixes for geos ≥ 3.8.0 (#224)

New in version 1.24

  • Fixed return value of arki-query (#221)

New in version 1.23

  • Added arki-query --progress
  • Restored processing.. message on arki-query --verbose (#216)
  • Added a periodic callback during queries, used to check for SIGINT and raising KeyboardError (#210)
  • Removed spurious returns in Summary.write (#214)
  • Fixed database queries with empty reftime values (#215)

New in version 1.22

  • Keep a Sesssion structure which tracks and reuses Segment objects, and as a consequence, file descriptors (#213)
  • Fixed typo in GRIB2 scanner and related GRIB2 test

New in version 1.21

  • Fixed sorted data queries in python bindings (#212)

New in version 1.20

  • Fixed hanging of arki-query when using postprocessor (#209)
  • Fixed arki-dump from stdin

New in version 1.19

  • Fixed level and timerange formatter for GRIB1 and GRIB2 (#207)

New in version 1.18

  • Ported documentation to sphinx
  • Fixed arki-server
  • arki-web-proxy command

New in version 1.17

  • Greatly enlarged Python API
  • New HTTP API
  • Redesigned and unified all the documentation (see https://arpa-simc.github.io/arkimet/)
  • Ported command line scripts to python
  • Removed unused gridspace querymacro
  • Removed unused --targetfile option
  • Removed unused --report option
  • Ported all Lua infrastructure and scripts to Python
  • Ported to new dballe 8 API
  • Cleaned up tests and autoconf checks
  • Use POSIX return values from command line tools (see https://www.freebsd.org/cgi/man.cgi?query=sysexits)
  • Work when sendfile is not available
  • ondisk2 datasets are deprecated from now on in favour of iseg, but still fully supported
  • sphinx-based documentation for python bindings
  • Do not stop at reporting the first 'dirty' cause found in a segment (#187)
  • Fixed order of iteration / overriding options of scanners (#177)
  • Fixed --annotate when some timedef keys are missing (#202)
  • Fixed errors in zip utils (#173)
  • Fixed bug on dataset compression (#170)
  • Fail a match on a metadata which does not have the data to match (#166)

New in version 1.12

  • Check and repack are able to fix an interrupted append to a dir dataset (#156)
  • Fixed writing VM2 as inline metadata (#160)
  • Fixed copyok/copyko empty file creation (#155)
  • Filter input from stdin (#164, #159)
  • Fixed tar option (#162)
  • Restored odim sequence file behaviour (#163)

New in version 1.9

  • implemented tar/zip segments (arki-check options --tar and --zip) (#131)
  • implemented soft bound on memory usage (default 128Mi, can be modified with --flush-threshold option in arki-scan, #152)
  • refactoring of stdin input: implemented --stdin (not available for --dispatch, #126)
  • grib definitions path now reflects environment variable if present
  • arki-check --remove now needs a --fix to actually delete data
  • Timerange handling for BUFR "generic" #(125)
  • fixed and improved tests (#128) (#139) (#135) (#140) (#151)
  • arki-xargs now removes temporary files if present (#124) (#134)
  • arki-check --state now look into archives (#95)
  • clarified move options scope in manpage (#146)
  • removed old files from source code (#133)

New in version 1.6

  • Locking is now disabled by default on file datasets and it's possible to disable it on dataset too using: locking=no option (#76)
  • Improved performance of arki-scan --dispatch for iseg and ondisk2 datasets (#110)
  • Arkimet can now be built on EcCodes (#121)
  • It's now possible to scan grib2 with different time units (#120)

Fixes

  • fixed yaml metadata parsing issue (#107)
  • fixed SQLite error with concurrent arki-check and arki-scan on iseg dataset (#115)
  • fixed ondisk2 and iseg datasets ignoring reftime matcher without date (#116)
  • fixed arki-query from metadata file (#117)
  • fixed race condition on simple dataset writes (#122)
  • fixed build without geos (#123)

New in version 1.5

  • Implemented single segment check on isegdatasets (#109)
  • Extended --filter option for all dataset types for --state, --repack and --remove-all operations (#111)
  • Fixed issue when importing a file with records extending across a number of steps greater than ulimit -n, arki-scan didn't import the data correctly, because the process has too many open files (#103)
  • Fixed concurrency issue on iseg datasets (#108)

New in version 1.4

  • Better HTTP code handling
  • Added documentation for datasets, segments, checks, fixes and repacks
  • Fixed some inconsistencies in arki-check behaviour
  • Test code refactoring
  • Added local arki-server for test suite
  • Other minor bug fixes

New in version 1.1

  • New dataset format with no dataset-wide index, and one .sqlite index per segment. The configuration required is: type = iseg format = grib|bufr|vm2|odimh5 (in order to be able to enumerate the segments of the right kind, they need to be told explicitly what data format they are going to store)

New in version 1.0

  • Removed dependency on libwibble

  • Minimal python3 bindings, and arki-server rewritten in python

  • Added arki-server --journald to format output messages so that they can be interpreted by journald

  • Added arki-server --perflog to log json info records for each query

  • Implemented sharded datasets: in a dataset configuration, use shard=yearly, shard=monthly, or shard=weekly, to have the dataset create subdatasets where data will be saved. Locking will only happen at the subdataset level.

    There is no automatic conversion from a non-sharded to a sharded dataset.

  • Implemented arki-check --unarchive=segment, to move a segment from .archives/last back to the main dataset.

  • Implemented arki-check --state to show the computed state of each segment in a dataset.

  • Implemented --copyok and --copyko options for arki-scan to copy somewhere every successfully imported (or failed) single message (i.e. a single grib in a multiple grib input file)

  • Removed arki-check --scantest

  • Simple and ondisk2 datasets now can only contain segments matching the dataset step. arki-check will give errors if it finds data files with a name that does not fit in the dataset.

    Archived data is not affected by this change.

  • Data scanning now only uses "grib" as a format for all grib editions, dropping the previous "grib1"/"grib2" distinction.

    This has the effect that newly created segments in datasets will have a .grib extension where previously they would have had a .grib1 or .grib2 extension. Existing segments are still supported.

  • AssignedDataset will not appear among new metadata anymore.