1
0
mirror of https://github.com/ilri/csv-metadata-quality.git synced 2024-12-14 16:32:19 +01:00

Add support for removing newlines

This was tricky because of the nature of newlines. In actuality we
are removing Unix line feeds here (U+000A) because Windows carriage
returns are actually already removed by the string stripping in the
whitespace fix.

Creating the test case in Vim was difficult because I couldn't fig-
ure out how to manually enter a line feed character. In the end I
used a search and replace on a known pattern like "ALAN", replacing
it with \r. Neither entering the Unicode code point (U+000A) direc-
tly or typing an "Enter" character after ^V worked. Grrr.
This commit is contained in:
Alan Orth 2019-07-30 20:05:12 +03:00
parent 346e66ca98
commit 40d5f7d81b
Signed by: alanorth
GPG Key ID: 0FB860CC9C45B1B9
5 changed files with 47 additions and 0 deletions

View File

@ -10,6 +10,7 @@ Requires Python 3.6 or greater. CSV and Excel support comes from the [Pandas](ht
- Validate subjects against the AGROVOC REST API
- Fix leading, trailing, and excessive (ie, more than one) whitespace
- Fix invalid multi-value separators (`|`) using `--unsafe-fixes`
- Fix problematic newlines (line feeds) using `--unsafe-fixes`
- Remove unnecessary Unicode like [non-breaking spaces](https://en.wikipedia.org/wiki/Non-breaking_space), [replacement characters](https://en.wikipedia.org/wiki/Specials_(Unicode_block)#Replacement_character), etc
- Check for "suspicious" characters that indicate encoding or copy/paste issues, for example "foreˆt" should be "forêt"
- Remove duplicate metadata values

View File

@ -25,6 +25,10 @@ def main(argv):
# Fix: whitespace
df[column] = df[column].apply(fix.whitespace)
# Fix: newlines
if args.unsafe_fixes:
df[column] = df[column].apply(fix.newlines)
# Fix: unnecessary Unicode
df[column] = df[column].apply(fix.unnecessary_unicode)

View File

@ -134,3 +134,34 @@ def duplicates(field):
new_field = '||'.join(new_values)
return new_field
def newlines(field):
"""Fix newlines.
Single metadata values should not span multiple lines because this is not
rendered properly in DSpace's XMLUI and even causes issues during import.
Implementation note: this currently only detects Unix line feeds (0x0a).
This is essentially when a user presses "Enter" to move to the next line.
Other newlines like the Windows carriage return are already handled with
the string stipping performed in the whitespace fixes.
Confusingly, in Vim '\n' matches a line feed when searching, but you must
use '\r' to *insert* a line feed, ie in a search and replace expression.
Return string with newlines removed.
"""
# Skip fields with missing values
if pd.isna(field):
return
# Check for Unix line feed (LF)
match = re.findall(r'\n', field)
if match:
print(f'Removing newline: {field}')
field = field.replace('\n', '')
return field

View File

@ -17,3 +17,5 @@ Invalid ISO 639-2 language,2019-07-29,,,jp,
Invalid ISO 639-3 language,2019-07-29,,,chi,
Invalid language,2019-07-29,,,Span,
Invalid AGROVOC subject,2019-07-29,,,,FOREST
Newline,2019-07-30,,,,"TANZA
NIA"

1 dc.contributor.author birthdate dc.identifier.issn dc.identifier.isbn dc.language.iso dc.subject
17 Invalid ISO 639-3 language 2019-07-29 chi
18 Invalid language 2019-07-29 Span
19 Invalid AGROVOC subject 2019-07-29 FOREST
20 Newline 2019-07-30 TANZA NIA
21

View File

@ -47,3 +47,12 @@ def test_fix_duplicates():
value = 'Kenya||Kenya'
assert fix.duplicates(value) == 'Kenya'
def test_fix_newlines():
'''Test fixing newlines.'''
value = '''Ken
ya'''
assert fix.newlines(value) == 'Kenya'