0.8.2-beta

What's Changed

• Add a step range filter parameter to methods that download series values

0.8.1

What's Changed

• Improved performance of ReadOnlyProject.fetch_*_df() by introducing concurrent data downloads

0.8.0

• when requesting data via fetch_(runs|experiments)_dfany run or experiment that matches the search criterion set by eg name_regex or custom_ids so on, will be returned regardless of the column filter
• limiting the set of columns by either providing a list (columns=[…]) or regex (columns_regex=’…’) is applied on top of the runs filtered out in the initial step
• those changes make the match_columns_to_filter argument obsolete and deprecated. We’ve kept it in the API and will remove it in a future release.
• Any column that is explicitly requested by the columns argument is guaranteed to exist in the resulting DataFrame, even if with only NA values
• • neptune-fetcher will now issue a warning when you go over a specified threshold of downloaded data cells, ie. rows * non-null-columns
    • The default is 1M, and can be changed using the NEPTUNE_WARN_AT_DATAFRAME_SIZE environment variable. Setting it to 0 disables the warning.
• The progress bar in fetch_*_df() is disabled for this release, and will be brought back in a subsequent version

0.7.0

Features

• Added support for initializing ReadOnlyRun based on experiment name (#54)
• Added fetch_read_only_experiments() method for fetching experiments in read-only mode (#54)

Changes

• Fixed columns_regex not respecting 5k column limit (#55)

0.6.0

Breaking Changes

• Removed sys/id from the columns that are always returned in fetch_*_df() (#51)
• Made passing None as columns parameter return only sys/custom_run_id (and sys/name in case of experiments) in fetch_*_df() (#51)

Changes

• Increased the limit of total columns fetched to 5000 in fetch_*_df() (#51)
• Moved regex matching for table filtering to the server side with NQL MATCHES operator in fetch_*_df() (#51)

Fixes

• Fixed bug where passing columns as None invalidated columns_regex parameter in fetch_*_df() (#51)
• Fixed bug where passing a names_regex not matching any experiment resulted in returning entire table (#51)
• Fixed misleading error message related to hitting a column number limit (#51)

Features

• Added match_columns_to_filters to conditionally enable column matching based on run filters (#51)

0.5.0

Features

• Added methods to list and fetch Experiments (#34)
• Added method to prefetch series values (#42)
• Added progress_bar parameter to prefetch_series_values (#43)
• Added support for filtering with Neptune Query Language (#45)
• Added support for include_inherited and progress_bar in series values fetching (#47)

0.5.0-rc0

Breaking Changes

• Removed names_regex parameter from fetch_runs_df() (#30)
• Removed sys/name from objects returned by run-related methods of ReadOnlyProject (#30)
• Made list_runs() and fetch_runs_df() methods return only Runs and not Experiments (#30)

Features

• Added methods to list and fetch Experiments (#34)
• Added method to prefetch series values (#42)

0.4.0

Features

• Added support for custom run IDs (#22)

0.3.0

Features

• Regex support with columns_regex and names_regex for fetch_runs_df() (#20)

0.2.0

Features

• Added support for bool, state, datetime and float series (#19)
• Added support for fetching float series values (#19)

Changes

• Using only paths filter endpoint instead of dedicated ones (#17)