Data Manipulation using Table¶
Section author: Gavin Huttley
The toolkit has a Table object that can be used for manipulating tabular data. It’s properties can be considered like an ordered 2 dimensional dictionary or tuple with flexible output format capabilities of use for exporting data for import into external applications. Importantly, via the restructured text format one can generate html or latex formatted tables. The table module is located within cogent3.util. The load_table and make_table convenience functions are provided as top-level cogent3 imports.
Table creation¶
Tables can be created directly using the Table object itself, or a convenience function that handles loading from files. We import both here:
>>> from cogent3 import load_table, make_table
>>> from cogent3.util.table import Table
First, if you try and create a Table without any data, it raises a ValueError.
>>> t = Table()
Traceback (most recent call last):
ValueError: header must be provided to Table
>>> t = Table(header=[], rows=[])
Traceback (most recent call last):
ValueError: header must be provided to Table
You can create a Table with no data, however.
>>> t = Table(header=["col 1", "col 2"], rows=[])
>>> t.shape == (0, 2)
True
Let’s create a very simple, rather nonsensical, table first. To create a table requires a header series, and a 2D series (either of type tuple, list, dict) or a pandas DataFrame..
>>> column_headings = ['chrom', 'stableid', 'length']
The string “chrom” will become the first column heading, “stableid” the second column heading, etc. The data are,
>>> rows = [['X', 'ENSG00000005893', 1353],
... ['A', 'ENSG00000019485', 1827],
... ['A', 'ENSG00000019102', 999],
... ['X', 'ENSG00000012174', 1599],
... ['X', 'ENSG00000010671', 1977],
... ['A', 'ENSG00000019186', 1554],
... ['A', 'ENSG00000019144', 4185],
... ['X', 'ENSG00000008056', 2307],
... ['A', 'ENSG00000018408', 1383],
... ['A', 'ENSG00000019169', 1698]]
...
>>>
We create the simplest of tables.
>>> t = Table(header=column_headings, rows=rows)
>>> print(t)
==================================
chrom stableid length
----------------------------------
X ENSG00000005893 1353
A ENSG00000019485 1827
A ENSG00000019102 999
X ENSG00000012174 1599
X ENSG00000010671 1977
A ENSG00000019186 1554
A ENSG00000019144 4185
X ENSG00000008056 2307
A ENSG00000018408 1383
A ENSG00000019169 1698
----------------------------------
The format above is referred to as ‘simple’ format in the documentation. Notice that the numbers in this table have 4 decimal places, despite the fact the original data were largely strings and had max of 3 decimal places precision. Table converts string representations of numbers to their appropriate form when you do str(table) or print the table.
We have several things we might want to specify when creating a table: the precision and or format of floating point numbers (integer argument - digits), the spacing between columns (integer argument or actual string of whitespace - space), title (argument - title), and legend (argument - legend). Lets modify some of these and provide a title and legend.
>>> t = Table(column_headings, rows, title='Alignment lengths',
... legend='Some analysis',
... digits=2, space=' ')
>>> print(t)
Alignment lengths
==========================================
chrom stableid length
------------------------------------------
X ENSG00000005893 1353
A ENSG00000019485 1827
A ENSG00000019102 999
X ENSG00000012174 1599
X ENSG00000010671 1977
A ENSG00000019186 1554
A ENSG00000019144 4185
X ENSG00000008056 2307
A ENSG00000018408 1383
A ENSG00000019169 1698
------------------------------------------
Some analysis
Note
The repr() of a table gives a quick summary.
>>> t
Alignment lengths
==========================================
chrom stableid length
------------------------------------------
X ENSG00000005893 1353
A ENSG00000019485 1827
A ENSG00000019102 999
X ENSG00000012174 1599
X ENSG00000010671 1977
A ENSG00000019186 1554
A ENSG00000019144 4185
X ENSG00000008056 2307
A ENSG00000018408 1383
A ENSG00000019169 1698
------------------------------------------
Some analysis
10 rows x 3 columns
The Table class cannot handle arbitrary python objects, unless they are passed in as strings. Note in this case we now directly pass in the column headings list and the handling of missing data can be explicitly specified..
>>> t2 = Table(['abcd', 'data'], [[str(list(range(1,6))), '0'],
... ['x', 5.0], ['y', None]],
... missing_data='*')
>>> print(t2)
=========================
abcd data
-------------------------
[1, 2, 3, 4, 5] 0
x 5.0000
y *
-------------------------
Table column headings can be assessed from the table.header property
>>> assert t2.header == ['abcd', 'data']
and this is immutable (cannot be changed).
>>> t2.header[1] = 'Data'
Traceback (most recent call last):
RuntimeError: Table header is immutable, use with_new_header
If you want to change the header, use the with_new_header method. This can be done one column at a time, or as a batch. The returned Table is identical aside from the modified column labels.
>>> mod_header = t2.with_new_header('abcd', 'ABCD')
>>> assert mod_header.header == ['ABCD', 'data']
>>> mod_header = t2.with_new_header(['abcd', 'data'], ['ABCD', 'DATA'])
>>> print(mod_header)
=========================
ABCD DATA
-------------------------
[1, 2, 3, 4, 5] 0
x 5.0000
y *
-------------------------
Tables may also be created from 2-dimensional dictionaries. In this case, special capabilities are provided to enforce printing rows in a particular order.
>>> d2D={'edge.parent': {'NineBande': 'root', 'edge.1': 'root',
... 'DogFaced': 'root', 'Human': 'edge.0', 'edge.0': 'edge.1',
... 'Mouse': 'edge.1', 'HowlerMon': 'edge.0'}, 'x': {'NineBande': 1.0,
... 'edge.1': 1.0, 'DogFaced': 1.0, 'Human': 1.0, 'edge.0': 1.0,
... 'Mouse': 1.0, 'HowlerMon': 1.0}, 'length': {'NineBande': 4.0,
... 'edge.1': 4.0, 'DogFaced': 4.0, 'Human': 4.0, 'edge.0': 4.0,
... 'Mouse': 4.0, 'HowlerMon': 4.0}, 'y': {'NineBande': 3.0, 'edge.1': 3.0,
... 'DogFaced': 3.0, 'Human': 3.0, 'edge.0': 3.0, 'Mouse': 3.0,
... 'HowlerMon': 3.0}, 'z': {'NineBande': 6.0, 'edge.1': 6.0,
... 'DogFaced': 6.0, 'Human': 6.0, 'edge.0': 6.0, 'Mouse': 6.0,
... 'HowlerMon': 6.0},
... 'edge.name': ['Human', 'HowlerMon', 'Mouse', 'NineBande', 'DogFaced',
... 'edge.0', 'edge.1']}
>>> row_order = d2D['edge.name']
>>> d2D['edge.name'] = dict(zip(row_order, row_order))
>>> t3 = Table(['edge.name', 'edge.parent', 'length', 'x', 'y', 'z'], d2D,
... row_order=row_order, missing_data='*', space=8,
... max_width=50, row_ids=True, title='My title',
... legend='legend: this is a nonsense example.')
>>> print(t3)
My title
==========================================
edge.name edge.parent length
------------------------------------------
Human edge.0 4.0000
HowlerMon edge.0 4.0000
Mouse edge.1 4.0000
NineBande root 4.0000
DogFaced root 4.0000
edge.0 edge.1 4.0000
edge.1 root 4.0000
------------------------------------------
continued: My title
=====================================
edge.name x y
-------------------------------------
Human 1.0000 3.0000
HowlerMon 1.0000 3.0000
Mouse 1.0000 3.0000
NineBande 1.0000 3.0000
DogFaced 1.0000 3.0000
edge.0 1.0000 3.0000
edge.1 1.0000 3.0000
-------------------------------------
continued: My title
=======================
edge.name z
-----------------------
Human 6.0000
HowlerMon 6.0000
Mouse 6.0000
NineBande 6.0000
DogFaced 6.0000
edge.0 6.0000
edge.1 6.0000
-----------------------
legend: this is a nonsense example.
In the above we specify a maximum width of the table, and also specify row identifiers (using row_ids, the integer corresponding to the column at which data begin, preceding columns are taken as the identifiers). This has the effect of forcing the table to wrap when the simple text format is used, but wrapping does not occur for any other format. The row_ids are keys for slicing the table by row, and as identifiers are presented in each wrapped sub-table.
Wrapping generate neat looking tables whether or not you index the table rows. We demonstrate here
>>> from cogent3 import make_table
>>> h = ['A/C', 'A/G', 'A/T', 'C/A']
>>> rows = [[0.0425, 0.1424, 0.0226, 0.0391]]
>>> wrap_table = make_table(header=h, rows=rows, max_width=30)
>>> print(wrap_table)
==========================
A/C A/G A/T
--------------------------
0.0425 0.1424 0.0226
--------------------------
continued:
======
C/A
------
0.0391
------
>>> wrap_table = make_table(header=h, rows=rows, max_width=30,
... row_ids=True)
>>> print(wrap_table)
==========================
A/C A/G A/T
--------------------------
0.0425 0.1424 0.0226
--------------------------
continued:
================
A/C C/A
----------------
0.0425 0.0391
----------------
We can also customise the formatting of individual columns.
>>> rows = (('NP_003077_hs_mm_rn_dna', 'Con', 2.5386013224378985),
... ('NP_004893_hs_mm_rn_dna', 'Con', 0.12135142635634111e+06),
... ('NP_005079_hs_mm_rn_dna', 'Con', 0.95165949788861326e+07),
... ('NP_005500_hs_mm_rn_dna', 'Con', 0.73827030202664901e-07),
... ('NP_055852_hs_mm_rn_dna', 'Con', 1.0933217708952725e+07))
We first create a table and show the default formatting behaviour for Table.
>>> t46 = Table(['Gene', 'Type', 'LR'], rows)
>>> print(t46)
===============================================
Gene Type LR
-----------------------------------------------
NP_003077_hs_mm_rn_dna Con 2.5386
NP_004893_hs_mm_rn_dna Con 121351.4264
NP_005079_hs_mm_rn_dna Con 9516594.9789
NP_005500_hs_mm_rn_dna Con 0.0000
NP_055852_hs_mm_rn_dna Con 10933217.7090
-----------------------------------------------
We then format the LR column to use a scientific number format.
>>> t46 = Table(['Gene', 'Type', 'LR'], rows)
>>> t46.format_column('LR', "%.4e")
>>> print(t46)
============================================
Gene Type LR
--------------------------------------------
NP_003077_hs_mm_rn_dna Con 2.5386e+00
NP_004893_hs_mm_rn_dna Con 1.2135e+05
NP_005079_hs_mm_rn_dna Con 9.5166e+06
NP_005500_hs_mm_rn_dna Con 7.3827e-08
NP_055852_hs_mm_rn_dna Con 1.0933e+07
--------------------------------------------
It is safe to directly modify certain attributes, such as the title, legend and white space separating columns, which we do for the t46.
>>> t46.title = "A new title"
>>> t46.legend = "A new legend"
>>> t46.space = ' '
>>> print(t46)
A new title
========================================
Gene Type LR
----------------------------------------
NP_003077_hs_mm_rn_dna Con 2.5386e+00
NP_004893_hs_mm_rn_dna Con 1.2135e+05
NP_005079_hs_mm_rn_dna Con 9.5166e+06
NP_005500_hs_mm_rn_dna Con 7.3827e-08
NP_055852_hs_mm_rn_dna Con 1.0933e+07
----------------------------------------
A new legend
We can provide settings for multiple columns.
>>> t3 = Table(['edge.name', 'edge.parent', 'length', 'x', 'y', 'z'], d2D,
... row_order=row_order)
>>> t3.format_column('x', "%.1e")
>>> t3.format_column('y', "%.2f")
>>> print(t3)
===============================================================
edge.name edge.parent length x y z
---------------------------------------------------------------
Human edge.0 4.0000 1.0e+00 3.00 6.0000
HowlerMon edge.0 4.0000 1.0e+00 3.00 6.0000
Mouse edge.1 4.0000 1.0e+00 3.00 6.0000
NineBande root 4.0000 1.0e+00 3.00 6.0000
DogFaced root 4.0000 1.0e+00 3.00 6.0000
edge.0 edge.1 4.0000 1.0e+00 3.00 6.0000
edge.1 root 4.0000 1.0e+00 3.00 6.0000
---------------------------------------------------------------
In some cases, the contents of a column can be of different types. In this instance, rather than passing a column template we pass a reference to a function that will handle this complexity. To illustrate this we will define a function that formats floating point numbers, but returns everything else as is.
>>> def formatcol(value):
... if isinstance(value, float):
... val = "%.2f" % value
... else:
... val = str(value)
... return val
We apply this to a table with mixed string, integer and floating point data.
>>> t6 = Table(['ColHead'], [['a'], [1], [0.3], ['cc']],
... column_templates=dict(ColHead=formatcol))
>>> print(t6)
=======
ColHead
-------
a
1
0.30
cc
-------
Creating a Table from a pandas DataFrame¶
Assign the DataFrame instance to the data_frame argument.
>>> from pandas import DataFrame
>>> df = DataFrame(data=[[0, 1], [3,7]], columns=['a', 'b'])
>>> print(df)
a b
0 0 1
1 3 7
>>> df_as_table = make_table(data_frame=df)
>>> print(df_as_table)
======
a b
------
0 1
3 7
------
Representation of tables¶
The representation formatting provides a quick overview of a table’s dimensions and it’s contents. We show this for a table with 3 columns and multiple rows
>>> t46
A new title
========================================
Gene Type LR
----------------------------------------
NP_003077_hs_mm_rn_dna Con 2.5386e+00
NP_004893_hs_mm_rn_dna Con 1.2135e+05
NP_005079_hs_mm_rn_dna Con 9.5166e+06
NP_005500_hs_mm_rn_dna Con 7.3827e-08
NP_055852_hs_mm_rn_dna Con 1.0933e+07
----------------------------------------
A new legend
5 rows x 3 columns
and larger
>>> t3
===============================================================
edge.name edge.parent length x y z
---------------------------------------------------------------
Human edge.0 4.0000 1.0e+00 3.00 6.0000
HowlerMon edge.0 4.0000 1.0e+00 3.00 6.0000
Mouse edge.1 4.0000 1.0e+00 3.00 6.0000
NineBande root 4.0000 1.0e+00 3.00 6.0000
DogFaced root 4.0000 1.0e+00 3.00 6.0000
edge.0 edge.1 4.0000 1.0e+00 3.00 6.0000
edge.1 root 4.0000 1.0e+00 3.00 6.0000
---------------------------------------------------------------
7 rows x 6 columns
Note
within a script use print(repr(t3)) to get the same representation.
Table output¶
Table can output in multiple formats, including restructured text or ‘rest’ and delimited. These can be obtained using the to_string method and format argument as follows. Using table t from above,
>>> print(t.to_string(format='rest'))
+----------------------------------+
| Alignment lengths |
+-------+-----------------+--------+
| chrom | stableid | length |
+=======+=================+========+
| X | ENSG00000005893 | 1353 |
+-------+-----------------+--------+
| A | ENSG00000019485 | 1827 |
+-------+-----------------+--------+
| A | ENSG00000019102 | 999 |
+-------+-----------------+--------+
| X | ENSG00000012174 | 1599 |
+-------+-----------------+--------+
| X | ENSG00000010671 | 1977 |
+-------+-----------------+--------+
| A | ENSG00000019186 | 1554 |
+-------+-----------------+--------+
| A | ENSG00000019144 | 4185 |
+-------+-----------------+--------+
| X | ENSG00000008056 | 2307 |
+-------+-----------------+--------+
| A | ENSG00000018408 | 1383 |
+-------+-----------------+--------+
| A | ENSG00000019169 | 1698 |
+-------+-----------------+--------+
| Some analysis |
+----------------------------------+
or Markdown format
>>> print(t.to_string(format='md'))
| chrom | stableid | length |
|-------|-----------------|--------|
| X | ENSG00000005893 | 1353 |
| A | ENSG00000019485 | 1827 |
| A | ENSG00000019102 | 999 |
| X | ENSG00000012174 | 1599 |...
which can also take an optional justify argument. The latter must be a series with a value for each column. (It only affects the html display of a Markdown table.)
>>> print(t.to_string(format='md', justify='lcr'))
| chrom | stableid | length |
|:------|:---------------:|-------:|
| X | ENSG00000005893 | 1353 |
| A | ENSG00000019485 | 1827 |
| A | ENSG00000019102 | 999 |
| X | ENSG00000012174 | 1599 |...
where the values lcr correspond to left, centre and right justification.
In the case of Markdown, the pipe character (|) is special and so cells containing it must be escaped.
>>> md_table = make_table(header=["a", "b"],
... rows=[["val1", "val2"],
... ["has | symbol", "val4"]])
>>> print(md_table.to_string(format='md'))
| a | b |
|---------------|------|
| val1 | val2 |
| has \| symbol | val4 |
Arguments such as space have no effect in this case. The table may also be written to file in any of the available formats (latex, simple text, html, pickle) or using a custom separator (such as a comma or tab). This makes it convenient to get data into other applications (such as R or a spreadsheet program).
The display format can be specified for a Table using any valid argument to to_string(). For instance, we can make a Table instance that defaults to Markdown display.
>>> md_table = make_table(header=["a", "b"],
... rows=[["val1", "val2"],
... ["has | symbol", "val4"]],
... format="md")
>>> print(md_table)
| a | b |
|---------------|------|
| val1 | val2 |
| has \| symbol | val4 |
This can be changed by modifying the format attribute, for example
>>> md_table.format = "rst"
>>> print(md_table)
+--------------+------+
| a | b |
+==============+======+
| val1 | val2 |
+--------------+------+
| has | symbol | val4 |
+--------------+------+
Here is the latex format, note how the title and legend are joined into the latex table caption. We also provide optional arguments for the column alignment (fist column left aligned, second column right aligned and remaining columns centred) and a label for table referencing.
>>> print(t3.to_string(format='tex', justify="lrcccc", label="table:example"))
\begin{table}[htp!]
\centering
\begin{tabular}{ l r c c c c }
\hline
\bf{edge.name} & \bf{edge.parent} & \bf{length} & \bf{x} & \bf{y} & \bf{z} \\
\hline
\hline
Human & edge.0 & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
HowlerMon & edge.0 & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
Mouse & edge.1 & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
NineBande & root & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
DogFaced & root & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
edge.0 & edge.1 & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
edge.1 & root & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
\hline
\end{tabular}
\label{table:example}
\end{table}
More complex latex table justifying is also possible. Specifying the width of individual columns requires passing in a series (list or tuple) of justification commands. In the following we introduce the command for specific columns widths.
>>> print(t3.to_string(format='tex', justify=["l","p{3cm}","c","c","c","c"]))
\begin{table}[htp!]
\centering
\begin{tabular}{ l p{3cm} c c c c }
\hline
\bf{edge.name} & \bf{edge.parent} & \bf{length} & \bf{x} & \bf{y} & \bf{z} \\
\hline
\hline
Human & edge.0 & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
HowlerMon & edge.0 & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
Mouse & edge.1 & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
NineBande & root & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
DogFaced & root & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
edge.0 & edge.1 & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
edge.1 & root & 4.0000 & 1.0e+00 & 3.00 & 6.0000 \\
\hline
\end{tabular}
\end{table}
>>> print(t3.to_string(sep=','))
edge.name,edge.parent,length, x, y, z
Human, edge.0,4.0000,1.0e+00,3.00,6.0000
HowlerMon, edge.0,4.0000,1.0e+00,3.00,6.0000
Mouse, edge.1,4.0000,1.0e+00,3.00,6.0000
NineBande, root,4.0000,1.0e+00,3.00,6.0000
DogFaced, root,4.0000,1.0e+00,3.00,6.0000
edge.0, edge.1,4.0000,1.0e+00,3.00,6.0000
edge.1, root,4.0000,1.0e+00,3.00,6.0000
You can specify any standard text character that will work with your desired target. Useful separators are tabs (\t), or pipes (|). If Table encounters the specified separator character within a cell, it wraps the cell in quotes – a standard approach to facilitate import by other applications. We will illustrate this with t2.
>>> print(t2.to_string(sep=', '))
abcd, data
"[1, 2, 3, 4, 5]", 0
x, 5.0000
y, *
Note that I introduced an extra space after the column just to make the result more readable in this example.
Test the writing of phylip distance matrix format.
>>> rows = [['a', '', 0.088337278874079342, 0.18848582712597683,
... 0.44084000179091454], ['c', 0.088337278874079342, '',
... 0.088337278874079342, 0.44083999937417828], ['b', 0.18848582712597683,
... 0.088337278874079342, '', 0.44084000179090932], ['e',
... 0.44084000179091454, 0.44083999937417828, 0.44084000179090932, '']]
>>> header = ['seq1/2', 'a', 'c', 'b', 'e']
>>> dist = Table(header=header, rows=rows, row_ids=True)
>>> print(dist.to_string(format='phylip'))
4
a 0.0000 0.0883 0.1885 0.4408
c 0.0883 0.0000 0.0883 0.4408
b 0.1885 0.0883 0.0000 0.4408
e 0.4408 0.4408 0.4408 0.0000
The to_string method also provides generic html generation via the restructured text format. The to_rich_html method can be used to generate the html table element by itself, with greater control over formatting. Specifically, users can provide custom callback functions to the row_cell_func and header_cell_func arguments to control in detail the formatting of table elements, or use the simpler dictionary based element_formatters approach. We use the above dist table to provide a specific callback that will set the background color for diagonal cells. We first write a function that takes the cell value and coordinates, returning the html formmatted text.
>>> def format_cell(value, row_num, col_num):
... bgcolor=['', ' bgcolor="#0055ff"'][value=='']
... return '<td%s>%s</td>' % (bgcolor, value)
We then call the method, without this argument, then with it.
>>> straight_html = dist.to_rich_html(compact=True)
>>> print(straight_html)
<table><thead style="font-weight: bold;"...
>>> rich_html = dist.to_rich_html(row_cell_func=format_cell,
... compact=False)
>>> print(rich_html)
<table>
<thead style="font-weight: bold;">
<th>seq1/2</th>
<th>a</th>
<th>c</th>
<th>b</th>
<th>e</th>
</thead>
<tbody>
<tr>
<td>a</td>
<td bgcolor="#0055ff"></td>
<td>0.0883</td>...
Convert Table to pandas DataFrame¶
If you have pandas installed, you can convert a Table instance to a DataFrame.
>>> tbl = Table(header=['a', 'b'], rows=[[0, 1], [3,7]])
>>> df = tbl.to_dataframe()
>>> type(df)
<class 'pandas.core.frame.DataFrame'>
>>> print(df)
a b
0 0 1
1 3 7
Exporting bedGraph format¶
One export format available is bedGraph. This format can be used for viewing data as annotation track in a genome browser. This format allows for unequal spans and merges adjacent spans with the same value. The format has many possible arguments that modify the appearance in the genome browser. For this example we just create a simple data set.
>>> rows = [['1', 100, 101, 1.123], ['1', 101, 102, 1.123],
... ['1', 102, 103, 1.123], ['1', 103, 104, 1.123],
... ['1', 104, 105, 1.123], ['1', 105, 106, 1.123],
... ['1', 106, 107, 1.123], ['1', 107, 108, 1.123],
... ['1', 108, 109, 1], ['1', 109, 110, 1],
... ['1', 110, 111, 1], ['1', 111, 112, 1],
... ['1', 112, 113, 1], ['1', 113, 114, 1],
... ['1', 114, 115, 1], ['1', 115, 116, 1],
... ['1', 116, 117, 1], ['1', 117, 118, 1],
... ['1', 118, 119, 2], ['1', 119, 120, 2],
... ['1', 120, 121, 2], ['1', 150, 151, 2],
... ['1', 151, 152, 2], ['1', 152, 153, 2],
... ['1', 153, 154, 2], ['1', 154, 155, 2],
... ['1', 155, 156, 2], ['1', 156, 157, 2],
... ['1', 157, 158, 2], ['1', 158, 159, 2],
... ['1', 159, 160, 2], ['1', 160, 161, 2]]
...
>>> bgraph = make_table(header=['chrom', 'start', 'end', 'value'],
... rows=rows)
...
>>> print(bgraph.to_string(format='bedgraph', name='test track',
... graphType='bar', description='test of bedgraph', color=(255,0,0)))
track type=bedGraph name="test track" description="test of bedgraph" color=255,0,0 graphType=bar
1 100 108 1.12
1 108 118 1.00
1 118 161 2.00
The bedgraph formatter defaults to rounding values to 2 decimal places. You can adjust that precision using the digits argument.
>>> print(bgraph.to_string(format='bedgraph', name='test track',
... graphType='bar', description='test of bedgraph', color=(255,0,0),
... digits=0))
track type=bedGraph name="test track" description="test of bedgraph" color=255,0,0 graphType=bar
1 100 118 1.00
1 118 161 2.00
Note
Writing files in bedgraph format is done using the write(format='bedgraph', name='test track', description='test of bedgraph', color=(255,0,0)).
Saving a table for reloading¶
Saving a table object to file for later reloading can be done using the standard write method and filename argument to the Table constructor, specifying any of the formats supported by to_string. The table loading will recreate a table from raw data located at filename. To illustrate this, we first write out the table t3 in pickle format, then the table t2 in a csv (comma separated values format). We then remove it’s header and write/reload as a tsv (tab separated values format).
>>> t3 = Table(['edge.name', 'edge.parent', 'length', 'x', 'y', 'z'], d2D,
... row_order=row_order, missing_data='*', space=8,
... max_width=50, row_ids=True, title='My title',
... legend='legend: this is a nonsense example.')
>>> t3.write("t3.pickle")
>>> t3_loaded = load_table("t3.pickle")
>>> print(t3_loaded)
My title
==========================================
edge.name edge.parent length
------------------------------------------
Human edge.0 4.0000
HowlerMon edge.0 4.0000
Mouse edge.1 4.0000
NineBande root 4.0000
DogFaced root 4.0000
edge.0 edge.1 4.0000
edge.1 root 4.0000
------------------------------------------
continued: My title
=====================================
edge.name x y
-------------------------------------
Human 1.0000 3.0000
HowlerMon 1.0000 3.0000
Mouse 1.0000 3.0000
NineBande 1.0000 3.0000
DogFaced 1.0000 3.0000
edge.0 1.0000 3.0000
edge.1 1.0000 3.0000
-------------------------------------
continued: My title
=======================
edge.name z
-----------------------
Human 6.0000
HowlerMon 6.0000
Mouse 6.0000
NineBande 6.0000
DogFaced 6.0000
edge.0 6.0000
edge.1 6.0000
-----------------------
legend: this is a nonsense example.
>>> t2 = Table(['abcd', 'data'], [[str([1, 2, 3, 4, 5]), '0'], ['x', 5.0],
... ['y', None]], missing_data='*', title='A \ntitle')
>>> t2.write('t2.csv')
>>> t2_loaded = load_table('t2.csv', header=True, with_title=True)
>>> print(t2_loaded)
A
title
=========================
abcd data
-------------------------
[1, 2, 3, 4, 5] 0
x 5.0000
y
-------------------------
>>> t2.title = ""
>>> t2.write("t2.tsv")
>>> t2_loaded = load_table('t2.tsv')
>>> print(t2_loaded)
=========================
abcd data
-------------------------
[1, 2, 3, 4, 5] 0
x 5.0000
y
-------------------------
Note the missing_data attribute is not saved in the delimited format, but is in the pickle format. In the next case, I’m going to override the digits format on reloading of the table.
>>> t2 = Table(['abcd', 'data'], [[str([1, 2, 3, 4, 5]), '0'], ['x', 5.0],
... ['y', None]], missing_data='*', title='A \ntitle',
... legend="And\na legend too")
>>> t2.write('t2.csv', sep=',')
>>> t2_loaded = load_table('t2.csv', header=True, with_title=True,
... with_legend=True, sep=',', digits = 2)
>>> print(t2_loaded)
A
title
=======================
abcd data
-----------------------
[1, 2, 3, 4, 5] 0
x 5.00
y
-----------------------
And
a legend too
A few things to note about the delimited file saving: formatting arguments are lost in saving to a delimited format; the header argument specifies whether the first line of the file should be treated as the header; the with_title and with_legend arguments are necessary if the file contains them, otherwise they become the header or part of the table. Importantly, if you wish to preserve numerical precision use the pickle format.
pickle can load a useful object from the pickled Table by itself, without needing to know anything about the Table class.
>>> import pickle
>>> f = open("t3.pickle", "rb")
>>> pickled = pickle.load(f)
>>> f.close()
>>> sorted(pickled.keys())
['digits', 'format', 'header', 'legend', 'max_width', 'missing_data',...
>>> pickled['rows'][0]
['Human', 'edge.0', 4.0, 1.0, 3.0, 6.0]
We can read in a delimited format using a custom reader. There are two approaches. The first one allows specifying different type conversions for different columns. The second allows specifying a whole line-based parser.
You can also read and write tables in gzip compressed format. This can be done simply by ending a filename with ‘.gz’ or specifying compress=True. We write a compressed file the two different ways and read it back in.
>>> t2.write('t2.csv.gz', sep=',')
>>> t2_gz = load_table('t2.csv.gz', sep=',', with_title=True,
... with_legend=True)
>>> t2_gz.shape == t2.shape
True
>>> t2.write('t2.csv', sep=',', compress=True)
>>> t2_gz = load_table('t2.csv.gz', sep=',', with_title=True,
... with_legend=True)
>>> t2_gz.shape == t2.shape
True
Defining a custom reader with type conversion for each column¶
We convert columns 2-5 to floats by specifying a field convertor. We then create a reader, specifying the data (below a list but can be a file) properties. Note that if no convertor is provided all data are returned as strings. We can also provide this reader to the Table constructor for a more direct way of opening such files. In this case, Table assumes there is a header row and nothing else.
>>> from cogent3.parse.table import ConvertFields, SeparatorFormatParser
>>> t3.title = t3.legend = None
>>> comma_sep = t3.to_string(sep=",").splitlines()
>>> print(comma_sep)
['edge.name,edge.parent,length, x, y, z', ' Human, ...
>>> converter = ConvertFields([(2,float), (3,float), (4,float), (5, float)])
>>> reader = SeparatorFormatParser(with_header=True,converter=converter,
... sep=",")
>>> comma_sep = [line for line in reader(comma_sep)]
>>> print(comma_sep)
[['edge.name', 'edge.parent', 'length', 'x', 'y', 'z'], ['Human',...
>>> t3.write("t3.tab", sep="\t")
>>> reader = SeparatorFormatParser(with_header=True,converter=converter,
... sep="\t")
>>> t3a = load_table("t3.tab", reader=reader, title="new title",
... space=2)
...
>>> print(t3a)
new title
======================================================
edge.name edge.parent length x y z
------------------------------------------------------
Human edge.0 4.0000 1.0000 3.0000 6.0000
HowlerMon edge.0 4.0000 1.0000 3.0000 6.0000
Mouse edge.1 4.0000 1.0000 3.0000 6.0000
NineBande root 4.0000 1.0000 3.0000 6.0000
DogFaced root 4.0000 1.0000 3.0000 6.0000
edge.0 edge.1 4.0000 1.0000 3.0000 6.0000
edge.1 root 4.0000 1.0000 3.0000 6.0000
------------------------------------------------------
We can use the SeparatorFormatParser to ignore reading certain lines by using a callback function. We illustrate this using the above data, skipping any rows with edge.name starting with edge.
>>> def ignore_internal_nodes(line):
... return line[0].startswith('edge')
...
>>> reader = SeparatorFormatParser(with_header=True,converter=converter,
... sep="\t", ignore=ignore_internal_nodes)
...
>>> tips = load_table("t3.tab", reader=reader, digits=1, space=2)
>>> print(tips)
=============================================
edge.name edge.parent length x y z
---------------------------------------------
Human edge.0 4.0 1.0 3.0 6.0
HowlerMon edge.0 4.0 1.0 3.0 6.0
Mouse edge.1 4.0 1.0 3.0 6.0
NineBande root 4.0 1.0 3.0 6.0
DogFaced root 4.0 1.0 3.0 6.0
---------------------------------------------
We can also limit the amount of data to be read in, very handy for checking large files.
>>> t3a = load_table("t3.tab", sep='\t', limit=3)
>>> print(t3a)
================================================================
edge.name edge.parent length x y z
----------------------------------------------------------------
Human edge.0 4.0000 1.0000 3.0000 6.0000
HowlerMon edge.0 4.0000 1.0000 3.0000 6.0000
Mouse edge.1 4.0000 1.0000 3.0000 6.0000
----------------------------------------------------------------
Limiting should also work when static_column_types is invoked
>>> t3a = load_table("t3.tab", sep='\t', limit=3, static_column_types=True)
>>> t3a.shape[0] == 3
True
or when
In the above example, the data type in a column is static, e.g. all values in x are floats. Rather than providing a custom reader, you can get the Table to construct such a reader based on the first data row using the static_column_types argument.
>>> t3a = load_table("t3.tab", static_column_types=True, digits=1,
... sep='\t')
>>> print(t3a)
=======================================================
edge.name edge.parent length x y z
-------------------------------------------------------
Human edge.0 4.0 1.0 3.0 6.0
HowlerMon edge.0 4.0 1.0 3.0 6.0
Mouse edge.1 4.0 1.0 3.0 6.0
NineBande root 4.0 1.0 3.0 6.0
DogFaced root 4.0 1.0 3.0 6.0
edge.0 edge.1 4.0 1.0 3.0 6.0
edge.1 root 4.0 1.0 3.0 6.0
-------------------------------------------------------
If you invoke the static_column_types argument and the column data are not static, you’ll get a ValueError. We show this by first creating a simple table with mixed data types in a column, write to file and then try to load with static_column_types=True.
>>> t3b = make_table(header=['A', 'B'], rows=[[1,1], ['a', 2]])
>>> print(t3b)
======
A B
------
1 1
a 2
------
>>> t3b.write('test3b.txt', sep='\t')
>>> t3b = load_table('test3b.txt', sep='\t', static_column_types=True)
Traceback (most recent call last):
ValueError: invalid literal for int() with base 10: 'a'
We also test the reader function for a tab delimited format with missing data at the end.
>>> data = ['ab\tcd\t', 'ab\tcd\tef']
>>> tab_reader = SeparatorFormatParser(sep='\t')
>>> for line in tab_reader(data):
... assert len(line) == 3, line
Defining a custom reader that operates on entire lines¶
It can also be the case that data types differ between lines. The basic mechanism is the same as above, except in defining the converter you must set the argument by_column=True.
We illustrate this capability by writing a short function that tries to cast entire lines to int, float or leaves as a string.
>>> def CastLine():
... floats = lambda x: list(map(float, x))
... ints = lambda x: list(map(int, x))
... def call(line):
... try:
... line = ints(line)
... except ValueError:
... try:
... line = floats(line)
... except ValueError:
... pass
... return line
... return call
We then define a couple of lines, create an instance of ConvertFields and call it for each type.
>>> line_str_ints = '\t'.join(map(str, range(5)))
>>> line_str_floats = '\t'.join(map(str, map(float, range(5))))
>>> data = [line_str_ints, line_str_floats]
>>> cv = ConvertFields(CastLine(), by_column=False)
>>> tab_reader = SeparatorFormatParser(with_header=False, converter=cv,
... sep='\t')
>>> for line in tab_reader(data):
... print(line)
[0, 1, 2, 3, 4]
[0.0, 1.0, 2.0, 3.0, 4.0]
Defining a custom writer¶
We can likewise specify a writer, using a custom field formatter and provide this to the Table directly for writing. We first illustrate how the writer works to generate output. We then use it to escape some text fields in quotes. In order to read that back in, we define a custom reader that strips these quotes off.
>>> from cogent3.format.table import format_fields, separator_formatter
>>> formatter = format_fields([(0,'"%s"'), (1,'"%s"')])
>>> writer = separator_formatter(formatter=formatter, sep=" | ")
>>> for formatted in writer(comma_sep, has_header=True):
... print(formatted)
edge.name | edge.parent | length | x | y | z
"Human" | "edge.0" | 4.0 | 1.0 | 3.0 | 6.0
"HowlerMon" | "edge.0" | 4.0 | 1.0 | 3.0 | 6.0
"Mouse" | "edge.1" | 4.0 | 1.0 | 3.0 | 6.0
"NineBande" | "root" | 4.0 | 1.0 | 3.0 | 6.0
"DogFaced" | "root" | 4.0 | 1.0 | 3.0 | 6.0
"edge.0" | "edge.1" | 4.0 | 1.0 | 3.0 | 6.0
"edge.1" | "root" | 4.0 | 1.0 | 3.0 | 6.0
>>> t3.write(filename="t3.tab", writer=writer)
>>> strip = lambda x: x.replace('"', '')
>>> converter = ConvertFields([(0,strip), (1, strip)])
>>> reader = SeparatorFormatParser(with_header=True, converter=converter,
... sep="|", strip_wspace=True)
>>> t3a = load_table("t3.tab", reader=reader, title="new title",
... space=2)
>>> print(t3a)
new title
=============================================
edge.name edge.parent length x y z
---------------------------------------------
Human edge.0 4.0 1.0 3.0 6.0
HowlerMon edge.0 4.0 1.0 3.0 6.0
Mouse edge.1 4.0 1.0 3.0 6.0
NineBande root 4.0 1.0 3.0 6.0
DogFaced root 4.0 1.0 3.0 6.0
edge.0 edge.1 4.0 1.0 3.0 6.0
edge.1 root 4.0 1.0 3.0 6.0
---------------------------------------------
Note
There are performance issues for large files. Pickling has proven very slow for saving very large files and introduces significant file size bloat. A simple delimited format is much more efficient both storage wise and, if you use a custom reader (or specify static_column_types=True), to generate and read. A custom reader was approximately 6 fold faster than the standard delimited file reader.
Table slicing and iteration¶
The Table class is capable of slicing by row, range of rows, column or range of columns headings or used to identify a single cell. Slicing using the method get_columns can also be used to reorder columns. In the case of columns, either the string headings or their position integers can be used. For rows, if row_ids was specified as True the 0’th cell in each row can also be used.
>>> t4 = Table(['edge.name', 'edge.parent', 'length', 'x', 'y', 'z'], d2D,
... row_order=row_order, row_ids=True, title='My title')
We subset t4 by column and reorder them.
>>> new = t4.get_columns(['z', 'y'])
>>> print(new)
My title
=============================
edge.name z y
-----------------------------
Human 6.0000 3.0000
HowlerMon 6.0000 3.0000
Mouse 6.0000 3.0000
NineBande 6.0000 3.0000
DogFaced 6.0000 3.0000
edge.0 6.0000 3.0000
edge.1 6.0000 3.0000
-----------------------------
We use the column position indexes to do get the same table.
>>> new = t4.get_columns([5, 4])
>>> print(new)
My title
=============================
edge.name z y
-----------------------------
Human 6.0000 3.0000
HowlerMon 6.0000 3.0000
Mouse 6.0000 3.0000
NineBande 6.0000 3.0000
DogFaced 6.0000 3.0000
edge.0 6.0000 3.0000
edge.1 6.0000 3.0000
-----------------------------
We can also using more general slicing, by both rows and columns. The following returns all rows from 4 on, and columns up to (but excluding) ‘y’:
>>> k = t4[4:, :'y']
>>> print(k)
My title
============================================
edge.name edge.parent length x
--------------------------------------------
DogFaced root 4.0000 1.0000
edge.0 edge.1 4.0000 1.0000
edge.1 root 4.0000 1.0000
--------------------------------------------
We can explicitly reference individual cells, in this case using both row and column keys.
>>> val = t4['HowlerMon', 'y']
>>> print(val)
3.0
We slice a single row,
>>> new = t4[3]
>>> print(new)
My title
================================================================
edge.name edge.parent length x y z
----------------------------------------------------------------
NineBande root 4.0000 1.0000 3.0000 6.0000
----------------------------------------------------------------
and range of rows.
>>> new = t4[3:6]
>>> print(new)
My title
================================================================
edge.name edge.parent length x y z
----------------------------------------------------------------
NineBande root 4.0000 1.0000 3.0000 6.0000
DogFaced root 4.0000 1.0000 3.0000 6.0000
edge.0 edge.1 4.0000 1.0000 3.0000 6.0000
----------------------------------------------------------------
You can iterate over the table one row at a time and slice the rows. We illustrate this for slicing a single column,
>>> for row in t:
... print(row['stableid'])
ENSG00000005893
ENSG00000019485
ENSG00000019102...
and for multiple columns.
>>> for row in t:
... print(row['stableid'], row['length'])
ENSG00000005893 1353
ENSG00000019485 1827
ENSG00000019102 999...
The numerical slice equivalent to the first case above would be row[0], to the second case either row[:], row[:2].
Filtering tables - selecting subsets of rows/columns¶
We want to be able to slice a table, based on some condition(s), to produce a new subset table. For instance, we construct a table with type and probability values.
>>> header = ['Gene', 'type', 'LR', 'df', 'Prob']
>>> rows = (('NP_003077_hs_mm_rn_dna', 'Con', 2.5386, 1, 0.1111),
... ('NP_004893_hs_mm_rn_dna', 'Con', 0.1214, 1, 0.7276),
... ('NP_005079_hs_mm_rn_dna', 'Con', 0.9517, 1, 0.3293),
... ('NP_005500_hs_mm_rn_dna', 'Con', 0.7383, 1, 0.3902),
... ('NP_055852_hs_mm_rn_dna', 'Con', 0.0000, 1, 0.9997),
... ('NP_057012_hs_mm_rn_dna', 'Unco', 34.3081, 1, 0.0000),
... ('NP_061130_hs_mm_rn_dna', 'Unco', 3.7986, 1, 0.0513),
... ('NP_065168_hs_mm_rn_dna', 'Con', 89.9766, 1, 0.0000),
... ('NP_065396_hs_mm_rn_dna', 'Unco', 11.8912, 1, 0.0006),
... ('NP_109590_hs_mm_rn_dna', 'Con', 0.2121, 1, 0.6451),
... ('NP_116116_hs_mm_rn_dna', 'Unco', 9.7474, 1, 0.0018))
>>> t5 = Table(header, rows)
>>> print(t5)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_003077_hs_mm_rn_dna Con 2.5386 1 0.1111
NP_004893_hs_mm_rn_dna Con 0.1214 1 0.7276
NP_005079_hs_mm_rn_dna Con 0.9517 1 0.3293
NP_005500_hs_mm_rn_dna Con 0.7383 1 0.3902
NP_055852_hs_mm_rn_dna Con 0.0000 1 0.9997
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_061130_hs_mm_rn_dna Unco 3.7986 1 0.0513
NP_065168_hs_mm_rn_dna Con 89.9766 1 0.0000
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_109590_hs_mm_rn_dna Con 0.2121 1 0.6451
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
---------------------------------------------------------
We then seek to obtain only those rows that contain probabilities < 0.05. We use valid python code within a string. Note: Make sure your column headings could be valid python variable names or the string based approach will fail (you could use an external function instead, see below).
>>> sub_table1 = t5.filtered(callback="Prob < 0.05")
>>> print(sub_table1)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_065168_hs_mm_rn_dna Con 89.9766 1 0.0000
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
---------------------------------------------------------
Using the above table we test the function to extract the raw data for a single column,
>>> raw = sub_table1.tolist('LR')
>>> raw
[34.3081..., 89.9766..., 11.8912, 9.7474...]
and from multiple columns.
>>> raw = sub_table1.tolist(columns=['df', 'Prob'])
>>> raw
[[1, 0.0], [1, 0.0],...
We can also do filtering using an external function, in this case we use a lambda to obtain only those rows of type ‘Unco’ that contain probabilities < 0.05, modifying our callback function.
>>> sub_table2 = t5.filtered(
... lambda ty_pr: ty_pr[0] == 'Unco' and ty_pr[1] < 0.05,
... columns=('type', 'Prob')
... )
>>> print(sub_table2)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
---------------------------------------------------------
This can also be done using the string approach.
>>> sub_table2 = t5.filtered("type == 'Unco' and Prob < 0.05")
>>> print(sub_table2)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
---------------------------------------------------------
We can also filter table columns using filtered_by_column. Say we only want the numerical columns, we can write a callback that returns False if some numerical operation fails, True otherwise.
>>> def is_numeric(values):
... try:
... sum(values)
... except TypeError:
... return False
... return True
>>> print(t5.filtered_by_column(callback=is_numeric))
=======================
LR df Prob
-----------------------
2.5386 1 0.1111
0.1214 1 0.7276
0.9517 1 0.3293
0.7383 1 0.3902
0.0000 1 0.9997
34.3081 1 0.0000
3.7986 1 0.0513
89.9766 1 0.0000
11.8912 1 0.0006
0.2121 1 0.6451
9.7474 1 0.0018
-----------------------
Appending tables¶
Tables may also be appended to each other, to make larger tables. We’ll construct two simple tables to illustrate this.
>>> geneA = Table(['edge.name', 'edge.parent', 'z'], [['Human','root',
... 6.0],['Mouse','root', 6.0], ['Rat','root', 6.0]],
... title='Gene A')
>>> geneB = Table(['edge.name', 'edge.parent', 'z'], [['Human','root',
... 7.0],['Mouse','root', 7.0], ['Rat','root', 7.0]],
... title='Gene B')
>>> print(geneB)
Gene B
==================================
edge.name edge.parent z
----------------------------------
Human root 7.0000
Mouse root 7.0000
Rat root 7.0000
----------------------------------
we now use the appended Table method to create a new table, specifying that we want a new column created (by passing the new_column argument a heading) in which the table titles will be placed.
>>> new = geneA.appended('Gene', geneB, title='Appended tables')
>>> print(new)
Appended tables
============================================
Gene edge.name edge.parent z
--------------------------------------------
Gene A Human root 6.0000
Gene A Mouse root 6.0000
Gene A Rat root 6.0000
Gene B Human root 7.0000
Gene B Mouse root 7.0000
Gene B Rat root 7.0000
--------------------------------------------
We repeat this without adding a new column.
>>> new = geneA.appended(None, geneB, title="Appended, no new column")
>>> print(new)
Appended, no new column
==================================
edge.name edge.parent z
----------------------------------
Human root 6.0000
Mouse root 6.0000
Rat root 6.0000
Human root 7.0000
Mouse root 7.0000
Rat root 7.0000
----------------------------------
Miscellaneous¶
Tables have a shape attribute, which specifies x (number of columns) and y (number of rows). The attribute is a tuple and we illustrate it for the above sub_table tables. Combined with the filtered method, this attribute can tell you how many rows satisfy a specific condition.
>>> t5.shape
(11, 5)
>>> sub_table1.shape
(4, 5)
>>> sub_table2.shape
(3, 5)
For instance, 3 of the 11 rows in t were significant and belonged to the Unco type.
For completeness, we generate a table with no rows and assess its shape.
>>> sub_table3 = t5.filtered(
... lambda ty_pr: ty_pr[0] == 'Unco' and ty_pr[1] > 0.1,
... columns=('type', 'Prob'))
>>> sub_table3.shape
(0, 5)
The distinct values can be obtained for a single column,
>>> distinct = new.distinct_values("edge.name")
>>> assert distinct == set(['Rat', 'Mouse', 'Human'])
or multiple columns
>>> distinct = new.distinct_values(["edge.parent", "z"])
>>> assert distinct == set([('root', 6.0), ('root', 7.0)]), distinct
We can compute column sums. Assuming only numerical values in a column.
>>> assert new.summed('z') == 39., new.summed('z')
We construct an example with mixed numerical and non-numerical data. We now compute the column sum with mixed non-numerical/numerical data.
>>> mix = make_table(header=['A', 'B'], rows=[[0,''],[1,2],[3,4]])
>>> print(mix)
======
A B
------
0
1 2
3 4
------
>>> mix.summed('B', strict=False)
6
We also compute row sums for the pure numerical and mixed non-numerical/numerical rows. For summing across rows we must specify the actual row index as an int.
>>> mix.summed(0, col_sum=False, strict=False)
0
>>> mix.summed(1, col_sum=False)
3
We can compute the totals for all columns or rows too.
>>> mix.summed(strict=False)
[4, 6]
>>> mix.summed(col_sum=False, strict=False)
[0, 3, 7]
It is not currently possible to do a subset of columns/rows. We show this for rows here.
>>> mix.summed([0, 2], col_sum=False, strict=False)
Traceback (most recent call last):
RuntimeError: unknown indices type: [0, 2]
We test these for a strictly numerical table.
>>> non_mix = make_table(header=['A', 'B'], rows=[[0,1],[1,2],[3,4]])
>>> non_mix.summed()
[4, 7]
>>> non_mix.summed(col_sum=False)
[1, 3, 7]
We can normalise a numerical table by row,
>>> print(non_mix.normalized(by_row=True))
================
A B
----------------
0.0000 1.0000
0.3333 0.6667
0.4286 0.5714
----------------
or by column, such that the row/column sums are 1.
>>> print(non_mix.normalized(by_row=False))
================
A B
----------------
0.0000 0.1429
0.2500 0.2857
0.7500 0.5714
----------------
We normalize by an arbitrary function (maximum value) by row,
>>> print(non_mix.normalized(by_row=True, denominator_func=max))
================
A B
----------------
0.0000 1.0000
0.5000 1.0000
0.7500 1.0000
----------------
by column.
>>> print(non_mix.normalized(by_row=False, denominator_func=max))
================
A B
----------------
0.0000 0.2500
0.3333 0.5000
1.0000 1.0000
----------------
Extending tables¶
In some cases it is desirable to compute an additional column from existing column values. This is done using the with_new_column method. We’ll use t4 from above, adding two of the columns to create an additional column.
>>> t7 = t4.with_new_column('Sum', callback="z+x", digits=2)
>>> print(t7)
My title
==================================================================
edge.name edge.parent length x y z Sum
------------------------------------------------------------------
Human edge.0 4.00 1.00 3.00 6.00 7.00
HowlerMon edge.0 4.00 1.00 3.00 6.00 7.00
Mouse edge.1 4.00 1.00 3.00 6.00 7.00
NineBande root 4.00 1.00 3.00 6.00 7.00
DogFaced root 4.00 1.00 3.00 6.00 7.00
edge.0 edge.1 4.00 1.00 3.00 6.00 7.00
edge.1 root 4.00 1.00 3.00 6.00 7.00
------------------------------------------------------------------
We test this with an externally defined function.
>>> func = lambda x_y: x_y[0] * x_y[1]
>>> t7 = t4.with_new_column('Sum', callback=func, columns=("y","z"),
... digits=2)
>>> print(t7)
My title
===================================================================
edge.name edge.parent length x y z Sum
-------------------------------------------------------------------
Human edge.0 4.00 1.00 3.00 6.00 18.00
HowlerMon edge.0 4.00 1.00 3.00 6.00 18.00
Mouse edge.1 4.00 1.00 3.00 6.00 18.00
NineBande root 4.00 1.00 3.00 6.00 18.00
DogFaced root 4.00 1.00 3.00 6.00 18.00
edge.0 edge.1 4.00 1.00 3.00 6.00 18.00
edge.1 root 4.00 1.00 3.00 6.00 18.00
-------------------------------------------------------------------
>>> func = lambda x: x**3
>>> t7 = t4.with_new_column('Sum', callback=func, columns="y", digits=2)
>>> print(t7)
My title
===================================================================
edge.name edge.parent length x y z Sum
-------------------------------------------------------------------
Human edge.0 4.00 1.00 3.00 6.00 27.00
HowlerMon edge.0 4.00 1.00 3.00 6.00 27.00
Mouse edge.1 4.00 1.00 3.00 6.00 27.00
NineBande root 4.00 1.00 3.00 6.00 27.00
DogFaced root 4.00 1.00 3.00 6.00 27.00
edge.0 edge.1 4.00 1.00 3.00 6.00 27.00
edge.1 root 4.00 1.00 3.00 6.00 27.00
-------------------------------------------------------------------
Sorting tables¶
We want a table sorted according to values in a column.
>>> sorted = t5.sorted(columns='LR')
>>> print(sorted)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_055852_hs_mm_rn_dna Con 0.0000 1 0.9997
NP_004893_hs_mm_rn_dna Con 0.1214 1 0.7276
NP_109590_hs_mm_rn_dna Con 0.2121 1 0.6451
NP_005500_hs_mm_rn_dna Con 0.7383 1 0.3902
NP_005079_hs_mm_rn_dna Con 0.9517 1 0.3293
NP_003077_hs_mm_rn_dna Con 2.5386 1 0.1111
NP_061130_hs_mm_rn_dna Unco 3.7986 1 0.0513
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_065168_hs_mm_rn_dna Con 89.9766 1 0.0000
---------------------------------------------------------
We want a table sorted according to values in a subset of columns, note the order of columns determines the sort order.
>>> sorted = t5.sorted(columns=('LR', 'type'))
>>> print(sorted)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_055852_hs_mm_rn_dna Con 0.0000 1 0.9997
NP_004893_hs_mm_rn_dna Con 0.1214 1 0.7276
NP_109590_hs_mm_rn_dna Con 0.2121 1 0.6451
NP_005500_hs_mm_rn_dna Con 0.7383 1 0.3902
NP_005079_hs_mm_rn_dna Con 0.9517 1 0.3293
NP_003077_hs_mm_rn_dna Con 2.5386 1 0.1111
NP_061130_hs_mm_rn_dna Unco 3.7986 1 0.0513
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_065168_hs_mm_rn_dna Con 89.9766 1 0.0000
---------------------------------------------------------
We now do a sort based on 2 columns.
>>> sorted = t5.sorted(columns=('type', 'LR'))
>>> print(sorted)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_055852_hs_mm_rn_dna Con 0.0000 1 0.9997
NP_004893_hs_mm_rn_dna Con 0.1214 1 0.7276
NP_109590_hs_mm_rn_dna Con 0.2121 1 0.6451
NP_005500_hs_mm_rn_dna Con 0.7383 1 0.3902
NP_005079_hs_mm_rn_dna Con 0.9517 1 0.3293
NP_003077_hs_mm_rn_dna Con 2.5386 1 0.1111
NP_065168_hs_mm_rn_dna Con 89.9766 1 0.0000
NP_061130_hs_mm_rn_dna Unco 3.7986 1 0.0513
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
---------------------------------------------------------
Reverse sort a single column
>>> sorted = t5.sorted('LR', reverse='LR')
>>> print(sorted)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_065168_hs_mm_rn_dna Con 89.9766 1 0.0000
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
NP_061130_hs_mm_rn_dna Unco 3.7986 1 0.0513
NP_003077_hs_mm_rn_dna Con 2.5386 1 0.1111
NP_005079_hs_mm_rn_dna Con 0.9517 1 0.3293
NP_005500_hs_mm_rn_dna Con 0.7383 1 0.3902
NP_109590_hs_mm_rn_dna Con 0.2121 1 0.6451
NP_004893_hs_mm_rn_dna Con 0.1214 1 0.7276
NP_055852_hs_mm_rn_dna Con 0.0000 1 0.9997
---------------------------------------------------------
Sort by just specifying the reverse column
>>> sorted = t5.sorted(reverse='LR')
>>> print(sorted)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_065168_hs_mm_rn_dna Con 89.9766 1 0.0000
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
NP_061130_hs_mm_rn_dna Unco 3.7986 1 0.0513
NP_003077_hs_mm_rn_dna Con 2.5386 1 0.1111
NP_005079_hs_mm_rn_dna Con 0.9517 1 0.3293
NP_005500_hs_mm_rn_dna Con 0.7383 1 0.3902
NP_109590_hs_mm_rn_dna Con 0.2121 1 0.6451
NP_004893_hs_mm_rn_dna Con 0.1214 1 0.7276
NP_055852_hs_mm_rn_dna Con 0.0000 1 0.9997
---------------------------------------------------------
Reverse sort one column but not another
>>> sorted = t5.sorted(columns=('type', 'LR'), reverse='LR')
>>> print(sorted)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_065168_hs_mm_rn_dna Con 89.9766 1 0.0000
NP_003077_hs_mm_rn_dna Con 2.5386 1 0.1111
NP_005079_hs_mm_rn_dna Con 0.9517 1 0.3293
NP_005500_hs_mm_rn_dna Con 0.7383 1 0.3902
NP_109590_hs_mm_rn_dna Con 0.2121 1 0.6451
NP_004893_hs_mm_rn_dna Con 0.1214 1 0.7276
NP_055852_hs_mm_rn_dna Con 0.0000 1 0.9997
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
NP_061130_hs_mm_rn_dna Unco 3.7986 1 0.0513
---------------------------------------------------------
Reverse sort both columns.
>>> sorted = t5.sorted(columns=('type', 'LR'), reverse=('type', 'LR'))
>>> print(sorted)
=========================================================
Gene type LR df Prob
---------------------------------------------------------
NP_057012_hs_mm_rn_dna Unco 34.3081 1 0.0000
NP_065396_hs_mm_rn_dna Unco 11.8912 1 0.0006
NP_116116_hs_mm_rn_dna Unco 9.7474 1 0.0018
NP_061130_hs_mm_rn_dna Unco 3.7986 1 0.0513
NP_065168_hs_mm_rn_dna Con 89.9766 1 0.0000
NP_003077_hs_mm_rn_dna Con 2.5386 1 0.1111
NP_005079_hs_mm_rn_dna Con 0.9517 1 0.3293
NP_005500_hs_mm_rn_dna Con 0.7383 1 0.3902
NP_109590_hs_mm_rn_dna Con 0.2121 1 0.6451
NP_004893_hs_mm_rn_dna Con 0.1214 1 0.7276
NP_055852_hs_mm_rn_dna Con 0.0000 1 0.9997
---------------------------------------------------------
Joining Tables¶
The Table object is capable of joins or merging of records in two tables. There are two fundamental types of joins – inner and outer – with there being different sub-types. We demonstrate these first constructing some simple tables.
>>> a=Table(header=["index", "col2","col3"],
... rows=[[1,2,3],[2,3,1],[2,6,5]], title="A")
>>> print(a)
A
=====================
index col2 col3
---------------------
1 2 3
2 3 1
2 6 5
---------------------
>>> b=Table(header=["index", "col2","col3"],
... rows=[[1,2,3],[2,2,1],[3,6,3]], title="B")
>>> print(b)
B
=====================
index col2 col3
---------------------
1 2 3
2 2 1
3 6 3
---------------------
>>> c=Table(header=["index","col_c2"],rows=[[1,2],[3,2],[3,5]],title="C")
>>> print(c)
C
===============
index col_c2
---------------
1 2
3 2
3 5
---------------
For a natural inner join, only 1 copy of columns with the same name are retained. So we expect the headings to be identical between the table a/b and the result of a.joined(b) or b.joined(a).
>>> assert a.joined(b).header == b.header
>>> assert b.joined(a).header == a.header
For a standard inner join, the joined table should contain all columns from a and b excepting the index column(s). Simply providing a column name (or index) selects this behaviour. Note that in this case, column names from the second table are made unique by prefixing them with that tables title. If the right table does not have a title, a default value right is used.
>>> b.title = None
>>> c.joined(b)
===========================================
index col_c2 right_col2 right_col3
-------------------------------------------
1 2 2 3
3 2 6 3
3 5 6 3
-------------------------------------------
3 rows x 4 columns
>>> b.title = 'B'
>>> assert a.joined(b, "index").header == ["index", "col2", "col3",
... "B_col2", "B_col3"]
...
Note that the table title’s were used to prefix the column headings from the second table. We further test this using table c which has different dimensions.
>>> assert a.joined(c,"index").header == ["index","col2","col3",
... "C_col_c2"]
It’s also possible to specify index columns using numerical values, the results of which should be the same.
>>> assert a.joined(b,[0, 2]).tolist() ==\
... a.joined(b,["index","col3"]).tolist()
Additionally, it’s possible to provide two series of indices for the two tables. Here, they have identical values.
>>> assert a.joined(b, ["index", "col3"],["index", "col3"]).tolist()\
... == a.joined(b,["index","col3"]).tolist()
The results of a standard join between tables a and b are
>>> print(a.joined(b, ["index"], title='A&B'))
A&B
=========================================
index col2 col3 B_col2 B_col3
-----------------------------------------
1 2 3 2 3
2 3 1 2 1
2 6 5 2 1
-----------------------------------------
We demo the table specific indices.
>>> print(a.joined(c, ["col2"], ["index"], title='A&C by "col2/index"'))
A&C by "col2/index"
=================================
index col2 col3 C_col_c2
---------------------------------
2 3 1 2
2 3 1 5
---------------------------------
Tables a and c share a single row with the same value in the index column, hence a join by that index should return a table with just that row.
>>> print(a.joined(c, "index", title='A&C by "index"'))
A&C by "index"
=================================
index col2 col3 C_col_c2
---------------------------------
1 2 3 2
---------------------------------
A natural join of tables a and b results in a table with only rows that were identical between the two parents.
>>> print(a.joined(b, title='A&B Natural Join'))
A&B Natural Join
=====================
index col2 col3
---------------------
1 2 3
---------------------
We test the outer join by defining an additional table with different dimensions, and conducting a join specifying inner_join=False.
>>> d=Table(header=["index", "col_c2"], rows=[[5,42],[6,23]], title="D")
>>> print(d)
D
===============
index col_c2
---------------
5 42
6 23
---------------
>>> print(c.joined(d,inner_join=False, title='C&D Outer join'))
C&D Outer join
======================================
index col_c2 D_index D_col_c2
--------------------------------------
1 2 5 42
1 2 6 23
3 2 5 42
3 2 6 23
3 5 5 42
3 5 6 23
--------------------------------------
We establish the joined method works for mixtures of character and numerical data, setting some indices and some cell values to be strings.
>>> a=Table(header=["index", "col2","col3"],
... rows=[[1,2,"3"],["2",3,1],[2,6,5]], title="A")
>>> b=Table(header=["index", "col2","col3"],
... rows=[[1,2,"3"],["2",2,1],[3,6,3]], title="B")
>>> assert a.joined(b, ["index", "col3"],["index", "col3"]).tolist()\
... == a.joined(b,["index","col3"]).tolist()
We test that the joined method works when the column index orders differ.
>>> t1_header = ['a', 'b']
>>> t1_rows = [(1,2),(3,4)]
>>> t2_header = ['b', 'c']
>>> t2_rows = [(3,6),(4,8)]
>>> t1 = Table(t1_header, rows=t1_rows, title='t1')
>>> t2 = Table(t2_header, rows=t2_rows, title='t2')
>>> t3 = t1.joined(t2, columns_self=["b"], columns_other=["b"])
>>> print(t3)
==============
a b t2_c
--------------
3 4 8
--------------
We then establish that a join with no values does not cause a failure, just returns an empty Table.
>>> t4_header = ['b', 'c']
>>> t4_rows = [(5,6),(7,8)]
>>> t4 = make_table(header=t4_header, rows=t4_rows)
>>> t4.title = 't4'
>>> t5 = t1.joined(t4, columns_self=["b"], columns_other=["b"])
>>> print(t5)
==============
a b t4_c
--------------
--------------
Whose representation looks like
>>> t5
==============
a b t4_c
--------------
--------------
0 rows x 3 columns
Transposing a table¶
Tables can be transposed.
>>> from cogent3 import make_table
>>> title='#Full OTU Counts'
>>> header = ['#OTU ID', '14SK041', '14SK802']
>>> rows = [[-2920, '332', 294],
... [-1606, '302', 229],
... [-393, 141, 125],
... [-2109, 138, 120],
... [-5439, 104, 117],
... [-1834, 70, 75],
... [-18588, 65, 47],
... [-1350, 60, 113],
... [-2160, 57, 52],
... [-11632, 47, 36]]
>>> table = make_table(header=header,rows=rows,title=title)
>>> print(table)
#Full OTU Counts
=============================
#OTU ID 14SK041 14SK802
-----------------------------
-2920 332 294
-1606 302 229
-393 141 125
-2109 138 120
-5439 104 117
-1834 70 75
-18588 65 47
-1350 60 113
-2160 57 52
-11632 47 36
-----------------------------
We now transpose this. We require a new column heading for header data and an identifier for which existing column will become the header (default is index 0).
>>> tp = table.transposed(new_column_name='sample',
... select_as_header='#OTU ID', space=2)
...
>>> print(tp)
==============================================================================
sample -2920 -1606 -393 -2109 -5439 -1834 -18588 -1350 -2160 -11632
------------------------------------------------------------------------------
14SK041 332 302 141 138 104 70 65 60 57 47
14SK802 294 229 125 120 117 75 47 113 52 36
------------------------------------------------------------------------------
We test transposition with default value is the same.
>>> tp = table.transposed(new_column_name='sample', space=2)
...
>>> print(tp)
==============================================================================
sample -2920 -1606 -393 -2109 -5439 -1834 -18588 -1350 -2160 -11632
------------------------------------------------------------------------------
14SK041 332 302 141 138 104 70 65 60 57 47
14SK802 294 229 125 120 117 75 47 113 52 36
------------------------------------------------------------------------------
We test transposition selecting a different column to become the header.
>>> tp = table.transposed(new_column_name='sample',
... select_as_header='14SK802', space=2)
...
>>> print(tp)
==============================================================================
sample 294 229 125 120 117 75 47 113 52 36
------------------------------------------------------------------------------
#OTU ID -2920 -1606 -393 -2109 -5439 -1834 -18588 -1350 -2160 -11632
14SK041 332 302 141 138 104 70 65 60 57 47
------------------------------------------------------------------------------
Counting rows¶
We can count the number of rows for which a condition holds. This method uses the same arguments as filtered but returns an integer result only.
>>> print(c.count("col_c2 == 2"))
2
>>> print(c.joined(d,inner_join=False).count("index==3 and D_index==5"))
2
Testing a sub-component¶
Before using Table, we exercise some formatting code:
>>> from cogent3.format.table import formatted_cells, phylip_matrix, latex
We check we can format an arbitrary 2D list, without a header, using the formatted_cells function directly.
>>> data = [[230, 'acdef', 1.3], [6, 'cc', 1.9876]]
>>> head = ['one', 'two', 'three']
>>> header, formatted = formatted_cells(data, header=head)
>>> print(formatted)
[['230', 'acdef', '1.3000'], [' 6', ' cc', '1.9876']]
>>> print(header)
['one', ' two', ' three']
We directly test the latex formatting.
>>> print(latex(formatted, header, justify='lrl', caption='A legend',
... label="table:test"))
\begin{table}[htp!]
\centering
\begin{tabular}{ l r l }
\hline
\bf{one} & \bf{two} & \bf{three} \\
\hline
\hline
230 & acdef & 1.3000 \\
6 & cc & 1.9876 \\
\hline
\end{tabular}
\caption{A legend}
\label{table:test}
\end{table}