tablite 2022.7.7

multiprocessing enabled out-of-memory data analysis library for tabular data.

Tablite

---

Overview

Tablite 2022.7 features

Even smaller memory footprint. Tablite uses HDF5 as a backend with strong abstraction, so that copy/append/repetition of data is handled in pages. This is imperative for incremental data processing such as in the image on the right, where 43M rows are processed in 208 steps. Tablite achieves this by storing all data in /tmp/tablite.hdf5 so if your OS sits on SSD it will benefit from high IOPS, and permit slices of 9,000,000,000 rows in less than a second. Multiprocessing enabled by default Tablite has multiprocessing is implemented for bypassing the GIL on all major operations. CSV import is tested with 96M fields that are imported and type-mapped to native python types in 120 secs. All algorithms have been reworked to respect memory limits Tablite respects the limits of free memory by tagging the free memory and defining task size before each memory intensive task is initiated (join, groupby, data import, etc) 100% support for all python datatypes Tablite uses datatype mapping to HDF5 native types where possible and uses type mapping for non-native types such as timedelta, None, date, time… e.g. what you put in, is what you get out. Light weight - Tablite is ~200 kB.

click for full size image

Installation

Tablite

Install: pip install tablite
Usage: >>> from tablite import Table

General overview

A tablite Table is multiprocessing enabled by default and ...

• behaves like a dict with lists: my_table[column name] = [... data ...]
• handles all python datatypes natively: str, float, bool, int, date, datetime, time, timedelta and None
• uses HDF5 as storage which is faster than mmap'ed files for the average case. 10,000,000 integers python will use < 1 Mb RAM instead of 133.7 Mb (1M lists with 10 integers). The example below shows data from tests/test_filereader_time.py with 1 terabyte of data:

An instance of a table allows you to:

• get rows in a column as mytable['A']
• get rows across all columns as mytable[4:8]
• slice as mytable['A', 'B', slice(4,8) ].
• update individual values with mytable['A'][2] = new value
• update many values even faster with list comprehensions such as: mytable['A'] = [ f(x) for x in mytable['A'] if x % 2 != 0 ]

You can:

• Use Table.import_file to import csv*, tsv, txt, xls, xlsx, xlsm, ods, zip and logs. There is automatic type detection (see tutorial.ipynb)
• To peek into any supported file use get_headers which shows the first 10 rows.
• Use mytable.rows and mytable.columns to iterate over rows or columns.
• Create multi-key .index for quick lookups.
• Perform multi-key .sort,
• Filter using .any and .all to select specific rows.
• use multi-key .lookup and .join to find data across tables.
• Perform .groupby and reorganise data as a .pivot table with max, min, sum, first, last, count, unique, average, st.deviation, median and mode
• Append / concatenate tables with += which automatically sorts out the columns - even if they're not in perfect order.
• Should you tables be similar but not the identical you can use .stack to "stack" tables on top of each other.

You can store or send data using json, by:

• dumping to json: json_str = table.to_json(), or
• you can load it with Table.from_json(json_str).-

One-liners

• loop over rows: [ row for row in table.rows ]
• loop over columns: [ table[col_name] for col_name in table.columns ]
• slice: myslice = table['A', 'B', slice(0,None,15)]
• join: left_join = numbers.left_join(letters, left_keys=['colour'], right_keys=['color'], left_columns=['number'], right_columns=['letter'])
• lookup: travel_plan = friends.lookup(bustable, (DataTypes.time(21, 10), "<=", 'time'), ('stop', "==", 'stop'))
• groupby: group_by = table.groupby(keys=['C', 'B'], functions=[('A', gb.count)])
• pivot table my_pivot = t.pivot(rows=['C'], columns=['A'], functions=[('B', gb.sum), ('B', gb.count)], values_as_rows=False)
• index: indices = old_table.index(*old_table.columns)
• sort: lookup1_sorted = lookup_1.sort(**{'time': True, 'name':False, "sort_mode":'unix'})
• filter: true,false = unfiltered.filter( [{"column1": 'a', "criteria":">=", 'value2':3}, ... more criteria ... ], filter_type='all' )
• any: even = mytable.any('A': lambda x : x%2==0, 'B': lambda x > 0)
• all: even = mytable.all('A': lambda x : x%2==0, 'B': lambda x > 0)

Tutorial

To learn more see the tutorial.ipynb (Jupyter notebook)

Credits

• Martynas Kaunas - GroupBy functionality.
• Audrius Kulikajevas - Edge case testing / various bugs.