Releases: mgckind/easyaccess
Releases · mgckind/easyaccess
easyaccess 1.4.7
easyaccess 1.4.6
- No minor changes to the code, just the general documentation
- Adding Contribution guidelines and Coda of Conduct
- Adding Authors
- Adding FAQ
- Updating Readme and Installation instructions
- Reformatting code
- Update paper and references
easyaccess 1.4.5
- Restructuring code structure (PR #151)
- Add paper reference
- Minor fixes in unitests
- Update documentation
- Fix conda issues
easyaccess 1.4.4
- Improved version to work with public release DB
- Update dtypes to match Y3 tables (PR #149)
- Fix bytes to unicode issue for py3 when writing h5 files
- Fix bug with Oracle types NUMBER with no scale/digits
- Add desdr database and remove oldoper
- Fix HDF5 output files, (see #140)
- Add compression option to configuration, gzip is applied to .csv,.tab,.fits (latter only for 1 DB trip), bzip2 for hdf5
fits can be opened in append mode when they are compressed, workaround is to increase prefetch
easyaccess 1.4.3
easyaccess 1.4.2
easyaccess 1.4.1
easyaccess 1.4.0
- Refactor and huge clean up, now bin/ folder has the startup script (see PR #113 and #111)
- Fix metadata and cache lookup for autocompletion (see #92, #94, #93)
- Add estimate number of rows on describe_table when available (see #75)
- Toggle color mode without exiting interpreter (see PR #110)
- 80 character loading bar (see PR #103)
- Added more information to help command (see PR #100 and #109)
- Fix config option and boolean values (see Pr #115)
- Justify comments to the left in describe_table i(#105)
- Added
change_db
option switch between dabatases without logging out (see PR #90 , #86) - Fixes authentication and des service files
- Added api to access descut services directly (see PR #87)
- Fix readline issues related with some OS X (see #88 and #112)
easyaccess 1.3.1
- Adds optional purge to drop_table (see PR #74)
- Fix a bug setting prefetch parameters using config set #76
- Improve import functionality for inline queries (see PR #78)
- Add option to upload files in chunk of memory (in addition to the chunk by rows), using --memsize option. This calculates an approximate number of rows to upload on each trip, avoiding memory issues. (see PR #79)
easyaccess 1.3.0
- Fixes a bug with python2/3 compatibility (see Pull-request #36)
- Removes ; from query for python API (query_* functions) issue #37
- Added execproc to run sql/pl procedures, see help execproc for more information (see PR #48)
- Fix show_index and describe_table commands and queries (no more repeated rows) (see PR #42 and #43)
- Case insensitive options (PR #45)
- Reorganization of the structure of the code, this way is much simpler to develop (PR #49)
- Conda installation
- In-query python functions (beta) write your own function and call it from query#67
- Query and execution information in FITS header #50
- Fix bug when loading sql from file #55
- Dealing better with trailing white space #52
- Fix several minor issues: #37, #40
- configuration options at command line (reset and configuration per session) with --config
- Improved python API , added some extra functions (pandas_to_db) and imprived docs#68
- Load big files in chunks -- very useful to load big files in chunks without running into memory issues #66
- Fix find_tables (to include owners) and describe_table (data_type) commands
- Added unittest suites #65 #70
- Added --no_refresh option in command line to quick start up