bookdata/cli/bx/
extract.rs1use std::fs::File;
6use std::io::Write;
7use std::path::PathBuf;
8
9use zip::ZipArchive;
10
11use crate::prelude::*;
12
13#[derive(Args, Debug)]
14pub struct Extract {
15 #[arg(name = "ZIPFILE")]
17 zipfile: PathBuf,
18
19 #[arg(name = "OUTFILE")]
21 outfile: PathBuf,
22}
23
24impl Command for Extract {
25 fn exec(&self) -> Result<()> {
26 info!("reading {:?}", self.zipfile);
27 let file = File::open(&self.zipfile)?;
28 let mut zip = ZipArchive::new(file)?;
29 let mut entry = zip.by_name("BX-Book-Ratings.csv")?;
30 let mut data = entry.read_all_sized()?;
31
32 info!("cleaning up data file");
33
34 debug!("removing non-ASCII characters and carriage returns");
35 data.retain(|b| *b < 128 && *b != b'\r');
36
37 debug!("replacing semicolons to make CSV");
38 for i in 0..data.len() {
40 let c = data[i];
41 if c == b';' {
42 data[i] = b',';
43 }
44 }
45
46 debug!("splitting CSV header");
47 let data = String::from_utf8(data)?;
48 let pos = if let Some(p) = data.find('\n') {
49 p
50 } else {
51 error!("no newline found, corrupt input data?");
52 return Err(anyhow!("corrupt data"));
53 };
54 let (hdr, rest) = data.split_at(pos + 1);
55 if !hdr.starts_with("\"User-ID\",") {
56 error!("unexpected file header found");
57 info!("found header: “{}“", hdr);
58 info!("expected cleaned header to begin with “\"User-ID\",“");
59 return Err(anyhow!("corrupt data"));
60 }
61
62 info!("writing cleaned output");
63 let mut out = File::create(&self.outfile)?;
64 write!(out, "user,isbn,rating\n")?;
65 let csvin = csv::Reader::from_reader(rest.as_bytes());
66 for row in csvin.into_records() {
67 let row = row?;
68 let user = row.get(0).ok_or(anyhow!("invalid CSV row"))?;
69 let isbn = row.get(1).ok_or(anyhow!("invalid CSV row"))?;
70 let rating = row.get(2).ok_or(anyhow!("invalid CSV row"))?;
71
72 let mut isbn = isbn.to_uppercase();
73 isbn.retain(|c| (c >= '0' && c <= '9') || c == 'X');
74
75 if isbn.len() > 0 {
76 write!(out, "{},{},{}\n", user, isbn, rating)?;
77 }
78 }
79
80 Ok(())
81 }
82}