mirror of
https://git.lyx.org/repos/lyx.git
synced 2024-11-09 18:31:04 +00:00
04ecabef60
Remove support for Python 2 Use formatted strings where appropriated
224 lines
5.9 KiB
Python
224 lines
5.9 KiB
Python
# file csv2lyx.py
|
|
# This file is part of LyX, the document processor.
|
|
# Licence details can be found in the file COPYING.
|
|
|
|
# author Hartmut Haase
|
|
# author José Matos
|
|
# Full author contact details are available in file CREDITS
|
|
|
|
# This script reads a csv-table (file name.csv) and converts it into
|
|
# a LyX-table for versions 1.5.0 and higher (LyX table format 276).
|
|
# It uses Python's csv module for parsing.
|
|
# The original csv2lyx was witten by Antonio Gulino <antonio.gulino@tin.it>
|
|
# in Perl for LyX 1.x and modified for LyX table format 276 by the author.
|
|
#
|
|
import csv, unicodedata
|
|
import os, sys
|
|
import optparse
|
|
|
|
def error(message):
|
|
sys.stderr.write(message + '\n')
|
|
sys.exit(1)
|
|
|
|
header = """#csv2lyx created this file. For more info see http://www.lyx.org/
|
|
\\lyxformat 413
|
|
\\begin_document
|
|
\\begin_header
|
|
\\textclass article
|
|
\\use_default_options true
|
|
\\maintain_unincluded_children false
|
|
\\language english
|
|
\\language_package default
|
|
\\inputencoding auto
|
|
\\fontencoding global
|
|
\\font_roman default
|
|
\\font_sans default
|
|
\\font_typewriter default
|
|
\\font_default_family default
|
|
\\use_non_tex_fonts false
|
|
\\font_sc false
|
|
\\font_osf false
|
|
\\font_sf_scale 100
|
|
\\font_tt_scale 100
|
|
|
|
\\graphics default
|
|
\\default_output_format default
|
|
\\output_sync 0
|
|
\\bibtex_command default
|
|
\\index_command default
|
|
\\paperfontsize default
|
|
\\use_hyperref false
|
|
\\papersize default
|
|
\\use_geometry false
|
|
\\use_amsmath 1
|
|
\\use_esint 1
|
|
\\use_mhchem 1
|
|
\\use_mathdots 1
|
|
\\cite_engine basic
|
|
\\use_bibtopic false
|
|
\\use_indices false
|
|
\\paperorientation portrait
|
|
\\suppress_date false
|
|
\\use_refstyle 1
|
|
\\index Index
|
|
\\shortcut idx
|
|
\\color #008000
|
|
\\end_index
|
|
\\secnumdepth 3
|
|
\\tocdepth 3
|
|
\\paragraph_separation indent
|
|
\\paragraph_indentation default
|
|
\\quotes_language english
|
|
\\papercolumns 1
|
|
\\papersides 1
|
|
\\paperpagestyle default
|
|
\\tracking_changes false
|
|
\\output_changes false
|
|
\\html_math_output 0
|
|
\\html_css_as_file 0
|
|
\\html_be_strict false
|
|
\\end_header
|
|
|
|
\\begin_body
|
|
|
|
\\begin_layout Standard
|
|
\\begin_inset Tabular
|
|
<lyxtabular version="3" rows="%d" columns="%d">
|
|
<features tabularvalignment="middle">
|
|
"""
|
|
|
|
cell = """<cell alignment="left" valignment="top" usebox="none">
|
|
\\begin_inset Text
|
|
|
|
\\begin_layout Plain Layout
|
|
%s
|
|
\\end_layout
|
|
|
|
\\end_inset
|
|
</cell>"""
|
|
|
|
footer = """</lyxtabular>
|
|
|
|
\\end_inset
|
|
|
|
|
|
\\end_layout
|
|
|
|
\\end_body
|
|
\\end_document
|
|
"""
|
|
|
|
# processing command line options
|
|
# delegate this to standard module optparse
|
|
args = {}
|
|
args["usage"] = "Usage: csv2lyx [options] csvfile [file.lyx]"
|
|
|
|
args["description"] = """This script creates a LyX document containing a table created from a
|
|
comma-separated-value (CSV) file. The resulting LyX file can be opened
|
|
with LyX 1.5.0 or any later version.
|
|
If no options are given csv2lyx will try to infer the CSV type of the csvfile,
|
|
"""
|
|
parser = optparse.OptionParser(**args)
|
|
|
|
parser.set_defaults(excel ='', column_sep = '')
|
|
parser.add_option("-e", "--excel", metavar ="CHAR",
|
|
help = """CHAR corresponds to a CSV type:
|
|
'e': Excel-generated CSV file
|
|
't': Excel-generated TAB-delimited CSV file""")
|
|
parser.add_option("-s", "--separator", dest = "column_sep",
|
|
help = """column separator
|
|
't' means Tab""")
|
|
|
|
group = optparse.OptionGroup(parser, "Remarks", """If your CSV file contains special characters (e. g. umlauts,
|
|
accented letters, etc.) make sure it is coded in UTF-8 (unicode).
|
|
Else LyX will loose some cell contents. If your CSV file was not written according to the "Common Format and MIME Type for Comma-Separated Values (CSV) Files" (http://tools.ietf.org/html/rfc4180) there may be unexpected results.""")
|
|
parser.add_option_group(group)
|
|
|
|
(options, args) = parser.parse_args()
|
|
|
|
# validate input
|
|
if len(args) == 1:
|
|
infile = args[0]
|
|
fout = sys.stdout
|
|
elif len(args) == 2:
|
|
infile = args[0]
|
|
fout = open(args[1], 'w')
|
|
else:
|
|
parser.print_help()
|
|
sys.exit(1)
|
|
|
|
if not os.path.exists(infile):
|
|
error('File "%s" not found.' % infile)
|
|
|
|
dialects = {'' : None, 'e' : 'excel', 't' : 'excel-tab'}
|
|
if options.excel not in dialects:
|
|
parser.print_help()
|
|
sys.exit(1)
|
|
dialect = dialects[options.excel]
|
|
|
|
# Set Tab, if necessary
|
|
if options.column_sep == 't':
|
|
options.column_sep = "\t"
|
|
|
|
# when no special column separator is given, try to detect it:
|
|
if options.column_sep and dialect :
|
|
reader = csv.reader(open(infile), dialect = dialect, delimiter = options.column_sep)
|
|
else:
|
|
guesser = csv.Sniffer()
|
|
input_file = "".join(open(infile).readlines())
|
|
try:
|
|
dialect = guesser.sniff(input_file)
|
|
reader = csv.reader(open(infile), dialect = dialect)
|
|
except:
|
|
# older versions (python < 2.5) of csv have problems (bugs)
|
|
# that is why we try harder to get a result, this should work on most cases
|
|
# as it assumes that the separator is a comma (the c in csv :-) )
|
|
try:
|
|
reader = csv.reader(open(infile), dialect = dialect, delimiter = ',')
|
|
except:
|
|
reader = csv.reader(open(infile), delimiter = ',')
|
|
|
|
# read input
|
|
num_cols = 1 # max columns
|
|
rows = []
|
|
|
|
for row in reader:
|
|
num_cols = max(num_cols, len(row))
|
|
rows.append(row)
|
|
|
|
num_rows = len(rows) # number of lines
|
|
|
|
# create a LyX file
|
|
#####################
|
|
# write first part
|
|
####################
|
|
fout.write(header % (num_rows, num_cols))
|
|
|
|
#####################
|
|
# write table
|
|
####################
|
|
for i in range(num_cols):
|
|
fout.write('<column alignment="left" valignment="top" width="0pt">\n')
|
|
|
|
for j in range(num_rows):
|
|
row = ['<row>']
|
|
|
|
############################
|
|
# write contents of one line
|
|
############################
|
|
for i in range(len(rows[j])):
|
|
row.append( cell % rows[j][i].replace('\\','\\backslash\n'))
|
|
|
|
# If row has less columns than num_cols fill with blank entries
|
|
for i in range(len(rows[j]), num_cols):
|
|
row.append(cell % " ")
|
|
|
|
fout.write("\n".join(row) + '\n</row>\n')
|
|
|
|
#####################
|
|
# write last part
|
|
####################
|
|
fout.write(footer)
|
|
# close the LyX file
|
|
fout.close()
|