Analyser des fichiers TSV et à largeur fixe

Au-delà du CSV : autres formats de fichiers

library(readr)

# readr's flat file readers:
# read_csv()    - comma-separated
# read_csv2()   - semicolon-separated (European)
# read_tsv()    - tab-separated
# read_delim()  - any delimiter
# read_fwf()    - fixed-width format
# read_lines()  - raw text, one line per element

cat('All readr functions return tibbles with the same interface!')

read_tsv() — Valeurs séparées par des tabulations

library(readr)

# Simulated TSV using \t as delimiter
tsv_data <- 'name\tscore\tcity
Alice Smith\t85\tNew York, NY
Bob Jones\t92\tLos Angeles, CA
Carol White\t78\tChicago, IL'

# Note: cities contain commas — TSV handles this cleanly
df <- read_tsv(tsv_data, show_col_types=FALSE)
print(df)

read_delim() — N’importe quel séparateur

library(readr)

# Pipe-delimited data
pipe_data <- 'ID|Name|Department|Salary
1|Alice Smith|Engineering|95000
2|Bob Jones|Marketing|72000
3|Carol White|HR|65000'

df <- read_delim(pipe_data, delim='|', show_col_types=FALSE)
print(df)

read_delim() pour les fichiers séparés par des points-virgules

library(readr)

# European-style: semicolon delimiter, comma decimal
european_data <- 'name;price;quantity
Widget;9,99;100
Gadget;24,50;50
Sprocket;4,75;200'

# Method 1: read_csv2() shortcut
df1 <- read_csv2(european_data, show_col_types=FALSE)
print(df1)

# Method 2: explicit read_delim
# read_delim(data, delim=';', locale=locale(decimal_mark=','))

read_fwf() — Format à largeur fixe

library(readr)

# Fixed-width data (positions matter!)
fwf_data <- 'Alice    085NYC      
Bob      092LA       
Carol    078Chicago  '

# Column widths: name=9, score=3, city=10
df <- read_fwf(
  fwf_data,
  col_positions = fwf_widths(
    widths = c(9, 3, 10),
    col_names = c('name','score','city')
  ),
  show_col_types = FALSE
)

print(df)

fwf_positions() pour les colonnes de début et de fin

library(readr)

# Another fixed-width example using start/end positions
fwf_data <- '001ALICE   02024-01-15
002BOB     01   2024-02-20
003CAROL   03   2024-03-10'

df <- read_fwf(
  fwf_data,
  col_positions = fwf_positions(
    start = c(1, 4, 11, 14),
    end   = c(3, 10, 12, 23),
    col_names = c('id','name','score','date')
  ),
  show_col_types = FALSE
)

print(df)

fwf_empty() — Détecter automatiquement la largeur

library(readr)

# Well-spaced fixed-width data
fwf_data <- 'name     score  grade
Alice    85     B
Bob      92     A
Carol    78     C
Dave     88     B'

# Auto-detect column positions from whitespace
df <- read_fwf(
  fwf_data,
  col_positions = fwf_empty(fwf_data, col_names=c('name','score','grade')),
  show_col_types = FALSE
)

print(df)

read_lines() — Analyse de texte brut

library(readr)

# Read lines and filter manually
raw_text <- '# Comment line
# Another comment
name,score
Alice,85
Bob,92'

lines <- read_lines(raw_text)
print(lines)

# Filter out comment lines
clean_lines <- lines[!startsWith(lines, '#')]
df <- read_csv(paste(clean_lines, collapse='\n'), show_col_types=FALSE)
print(df)

read_lines() pour une inspection progressive

library(readr)

# Preview first few lines before reading
sample_file <- 'METADATA: version 1.2
DATE: 2024-01-15
---
id,name,value
1,Alice,100
2,Bob,200'

# Inspect to find where the data starts
preview <- read_lines(sample_file, n_max=4)
print(preview)
# From this we know: skip=3, then read as CSV

Gérer les problèmes d’encodage

library(readr)

# Detect encoding of a file
# readr::guess_encoding('file.csv')  # Returns likely encodings

# Specify encoding explicitly
# df <- read_csv('data_latin1.csv',
#   locale = locale(encoding = 'latin1'))

# For UTF-8 with BOM (common in Windows exports):
# df <- read_csv('data_utf8bom.csv',
#   locale = locale(encoding = 'UTF-8-BOM'))

cat('Encoding matters for accented chars: \xc3\xa9 (UTF-8 for e-acute)\n')
print(guess_encoding(chartr('\xe9','e','caf\xe9')))

Choisir la fonction de lecture appropriée

library(readr)

# Quick comparison of delimiters
csv_data  <- 'a,b,c\n1,2,3'
tsv_data  <- 'a\tb\tc\n1\t2\t3'
pipe_data <- 'a|b|c\n1|2|3'

read_csv(csv_data, show_col_types=FALSE)
read_tsv(tsv_data, show_col_types=FALSE)
read_delim(pipe_data, delim='|', show_col_types=FALSE)

Vérification rapide

Récapitulatif : fichiers TSV et à largeur fixe

library(readr)

# Pipe-delimited with custom NA values
df <- read_delim(
  'ID|Name|Score|City
  1|Alice|85|NYC
  2|Bob|N/A|N/A
  3|Carol|78|Chicago',
  delim = '|',
  na = c('', 'NA', 'N/A'),
  show_col_types = FALSE
)

print(df)
print(colSums(is.na(df)))