# Source formatting of Markdown Table

I use MultiMarkdown tables more than I thought I ever would. In my .md files, I would like the source table code to be neat, evenly spaced, and resemble (in an alignment sense) the final HTML rendered table.

## Solution

Select the source of the table, press a hotkey on my computer, replace the untidy table with the tidy table.

## Notes:

• This script (macro?) is for my own use; I will always write the source table the same way
• The script is written for Python 2.7
• I use a Mac app that allows me to assign a hot key to an action (e.g. a script), and the following seems to work with that at the moment!

### Example

I write something like the following:

| Header 1 | Header 2 | ... | Header m |
| :--- | :--: | :--: | --: |
| $a_{11}$ | $a_{12}$ | ... | $a_{1m}$ |
| ... | ... | ... | ... |
| $a_{n1}$ | $a_{n2}$ | ... | $a_{nm} |  select it, press my hot key, and it is replaced by something like this (in the source): | Header 1 | Header 2 | ... | Header m | | :--- | :--: | :--: | --: | |$a_{11}$|$a_{12}$| ... |$a_{1m}$| | ... | ... | ... | ... | |$a_{n1}$|$a_{n2}$| ... |$a_{nm}\$ |


As mentioned, the script seems to work.

How can I improve the script?

I would really appreciate some criticism of the script I've written. I am not a coder, but try on occasion to write little tool type things like this.

## The Script

#!/usr/bin/env python

"""MMD Table Formatter.

Silly script that takes a MMD table as a
string and returns a tidied version of the table
"""

import sys
import StringIO

query = sys.argv[1]

# For "cleaned" table entries:
rows = []

# This NEEDS TO BE CLOSED AFTER!!
s_obj = StringIO.StringIO(query)
# Clean the entries:
for line in s_obj:
l = line.split('|')
rows.append([entry.strip() for entry in l])enter code here
# CLOSE
s_obj.close()

# Max length of each "entry" is what we'll use
# to evenly space "columns" in the final table
cols = zip(*rows)
col_widths = []
for columns in cols:
row_widths = map(lambda x: len(x), columns)
col_widths.append(max(row_widths))

# Let's align entries as per intended formatting.
# Second line of input string contains alignment commmands:
# ":---"    left aligned
# "---:"    right aligned
# ":--:"    centered (also accepts "---")
alignment = []
for r in rows[1]:
if r.startswith(":") and not r.endswith(":"):
alignment.append("lalign")
elif r.endswith(":") and not r.startswith(":"):
alignment.append("ralign")
else:
alignment.append("centered")

# Prepare for output string:
out = []
for row in rows:
for entry_and_width in zip(row, col_widths, alignment):
if entry_and_width[1] == 0:
continue
if entry_and_width[2] == "centered":
outstring = "| " + entry_and_width[0].center(entry_and_width[1]) + ' '
out.append(outstring)
if entry_and_width[2] == "lalign":
outstring = "| " + entry_and_width[0].ljust(entry_and_width[1]) + ' '
out.append(outstring)
if entry_and_width[2] == "ralign":
outstring = "| " + entry_and_width[0].rjust(entry_and_width[1]) + ' '
out.append(outstring)
out.append("|\n")

query = "".join(out)

sys.stdout.write(query)

• Note that a similar question was asked a few days ago. You might get some insights there. – 409_Conflict Oct 26 '16 at 9:49

As far as I can see, there is no need for the StringIO (?). You can just use query.split('\n'). Regardless, that for loop can be condensed into a list comprehension:

rows = [[el.strip() for el in row.split('|')] for row in query.splitlines()]


If the StringIO is really needed, I would use with..as:

with StringIO.StringIO(query) as s_obj:
rows = [[el.strip() for el in row.split('|')] for row in s_obj]


For the column widths you can use that len is already a function, so there is no need for lambda x: len(x). This way you can also inline all of it into one list comprehension:

col_widths = [max(map(len, column)) for column in zip(*rows)]


For the alignments, I would define a function that returns the alignment, given the content of a cell. First I had this function return your strings. But then I realized that you are already using entry.ljust further down. I then changed it to return the function to use here. Note that str.ljust("ab", 2) and "ab".ljust(2) are equivalent, so later we just call align(entry, width).

def get_alignment(cell):
""""
:---"    left aligned
"---:"    right aligned
":--:"    centered (also accepts "---"), default
"""
if cell.startswith(":") and not cell.endswith(":"):
return str.ljust
elif cell.endswith(":") and not cell.startswith(":"):
return str.rjust
return str.center


To get all alignments, we just use map again:

alignments = map(get_alignment, rows[1])


Finally, the output part. Since you already build a nice tuple with zip, you should use tuple unpacking to give the element readable names. entry_and_widths is a confusing name, especially since it also contains the alignments!

Here we can now get rid of most of your code, since it boils down to:

entry = align(entry, width)
out.append("| {} ".format(entry)


Here I used str.format to make it a bit easier and avoid the costly string addition. Note that we could also use str.format to do the adjusting for us (using e.g. "{>2}".format(entry) instead of "{}".format(str.rjust(entry, 2)), but that would mean nesting formats, which starts to get ugly very quickly.

I also use the fact that 0 compares to False to make the code skipping a column if it is empty shorter.

out = []
for row in rows:
for entry, width, align in zip(row, col_widths, alignments):
if not width:
continue
out.append("| {} ".format(align(entry, width)))
out.append("|\n")
query = "".join(out)


Final code:

#!/usr/bin/env python

"""MMD Table Formatter.

Silly script that takes a MMD table as a
string and returns a tidied version of the table
"""
import sys

def get_alignment(cell):
""""
:---"    left aligned
"---:"    right aligned
":--:"    centered (also accepts "---"), default
"""
if cell.startswith(":") and not cell.endswith(":"):
return str.ljust
elif cell.endswith(":") and not cell.startswith(":"):
return str.rjust
return str.center

query = sys.argv[1]

# For "cleaned" table entries:
rows = [[el.strip() for el in row.split('|')] for row in query.splitlines()]

# Max length of each "entry" is what we'll use
# to evenly space "columns" in the final table
col_widths = [max(map(len, column)) for column in zip(*rows)]

# Let's align entries as per intended formatting.
# Second line of input string contains alignment commmands:
alignments = map(get_alignment, rows[1])

# Prepare for output string:
out = []
for row in rows:
for entry, width, align in zip(row, col_widths, alignments):
if not width:
continue
out.append("| {} ".format(align(entry, width)))
out.append("|\n")

query = "".join(out)

sys.stdout.write(query)

• You should use query.splitlines() rather than query.split('\n'). Applied to 'test string\n' the former yield ['test string'] whereas the latter yield ['test string', ''] which is not that convenient to work with. – 409_Conflict Oct 26 '16 at 9:52
• Returning str.ljust et al., was exactly what I was looking for, and unaware I could do; so cheers. Thanks for reminding me of map too :) @MathiasEttinger Originally I had used splitlines() but opted for StringIO for speed - no idea if this is justified. – c ss Oct 26 '16 at 10:18
• @css As a wild gues I would say its not justified. Unless you had performances issues, ran a profiler and saw that the bottleneck was splitlines. – 409_Conflict Oct 26 '16 at 10:20

To complete on @Graipher's answer, I would simplify the formating of the rows using list-comprehensions/generator expressions rather than building a list to feed into join:

query = '\n'.join(
'| {} |'.format(  # Build a row composed of an inner part between delimiters
' | '.join(align(entry, width)
for entry, width, align in zip(row, col_widths, alignments)))
for row in rows)