adding check for duplicated pool names in the input genotype file

jeffersonfparil · Jun 25, 2024 · fedd75a · fedd75a
1 parent 8331afd
commit fedd75a
Show file tree

Hide file tree

Showing 4 changed files with 86 additions and 20 deletions.
diff --git a/src/aldknni.rs b/src/aldknni.rs
@@ -1013,7 +1013,7 @@ pub fn impute_aldknni(
             Ok(x) => x,
             Err(_) => return Err(ImputefError{
                 code: 129,
-                message: "Error writing the output file using the write_tsv() method within impute_aldknni(): ".to_owned() + &out
+                message: "Error writing the output file using the write_tsv() method within impute_aldknni(): ".to_owned() + out
             })
         };
     Ok(out)

diff --git a/src/geno.rs b/src/geno.rs
@@ -426,7 +426,7 @@ pub fn load_geno<'a, 'b>(
             Err(_) => {
                 return Err(ImputefError {
                     code: 316,
-                    message: "Error reading the allele frequency table file: ".to_owned() + &fname,
+                    message: "Error reading the allele frequency table file: ".to_owned() + fname,
                 })
             }
         },
@@ -435,7 +435,7 @@ pub fn load_geno<'a, 'b>(
                 code: 317,
                 message: "Please check the format of the allele frequency table text file: "
                     .to_owned()
-                    + &fname,
+                    + fname,
             })
         }
     };
@@ -460,26 +460,48 @@ pub fn load_geno<'a, 'b>(
         true => (),
         false => return Err(ImputefError{
             code: 318,
-            message: "Error unable to properly parse the header line. Please make sure the allele frequency table file: ".to_owned() + &fname +" is separated by tabs, commas, or semi-colons."
+            message: "Error unable to properly parse the header line. Please make sure the allele frequency table file: ".to_owned() + fname +" is separated by tabs, commas, or semi-colons."
         })
     };
     let pool_names: Vec<String> = vec_header[3..vec_header.len()]
         .iter()
         .map(|&x| x.to_owned())
         .collect();
     let n = pool_names.len();
+    // Check for duplicated pool names
+    let mut unique_pool_names: Vec<String> = vec![];
+    for name_source in pool_names.iter() {
+        let mut duplicated = false;
+        for name_destination in unique_pool_names.iter() {
+            if name_source == name_destination {
+                duplicated = true;
+                break;
+            }
+        }
+        if !duplicated {
+            unique_pool_names.push(name_source.to_string())
+        }
+    }
+    if n > unique_pool_names.len() {
+        return Err(ImputefError {
+            code: 139,
+            message: "Error: there are duplicated pool names in file: ".to_owned()
+                + fname
+                + " in load_geno() function.",
+        });
+    }
     // If a single pool size was supplied then we are assuming the same sizes across all pools
     if filter_stats.pool_sizes.len() == 1 {
         filter_stats.pool_sizes = vec![filter_stats.pool_sizes[0]; n];
     }
     match filter_stats.pool_sizes.len() == n {
         true => (),
         false => return Err(ImputefError {
-            code: 319,
+            code: 320,
             message:
                 "Error in the number of pools and the pool sizes do not match in the input file: "
                     .to_owned()
-                    + &fname,
+                    + fname,
         }),
     };
     let file_geno = FileGeno {
@@ -492,11 +514,11 @@ pub fn load_geno<'a, 'b>(
     ) {
         Ok(x) => x,
         Err(_e) => return Err(ImputefError {
-            code: 320,
+            code: 321,
             message:
                 "Error parsing the genotype data (extracted from allele frequency table text file: "
                     .to_owned()
-                    + &fname
+                    + fname
                     + ") via convert_into_genotypes_and_phenotypes() method within impute().",
         }),
     };

diff --git a/src/sync.rs b/src/sync.rs
@@ -1262,7 +1262,7 @@ pub fn load_sync<'a, 'b>(
         Err(_) => {
             return Err(ImputefError {
                 code: 763,
-                message: "Error reading the input vcf file: ".to_owned() + &fname,
+                message: "Error reading the input vcf file: ".to_owned() + fname,
             })
         }
     };
@@ -1273,7 +1273,7 @@ pub fn load_sync<'a, 'b>(
             Err(_) => {
                 return Err(ImputefError {
                     code: 764,
-                    message: "Error reading the input sync file: ".to_owned() + &fname,
+                    message: "Error reading the input sync file: ".to_owned() + fname,
                 })
             }
         };
@@ -1300,11 +1300,33 @@ pub fn load_sync<'a, 'b>(
             return Err(ImputefError {
                 code: 765,
                 message: "Error reading the header line of the sync file: ".to_owned()
-                    + &fname
+                    + fname
                     + ". Please make sure the header line starts with '#chr'.",
             })
         }
     };
+    // Check for duplicated pool names
+    let mut unique_pool_names: Vec<String> = vec![];
+    for name_source in pool_names.iter() {
+        let mut duplicated = false;
+        for name_destination in unique_pool_names.iter() {
+            if name_source == name_destination {
+                duplicated = true;
+                break;
+            }
+        }
+        if !duplicated {
+            unique_pool_names.push(name_source.to_string())
+        }
+    }
+    if n > unique_pool_names.len() {
+        return Err(ImputefError {
+            code: 766,
+            message: "Error: there are duplicated pool names in file: ".to_owned()
+                + fname
+                + " in load_sync() function.",
+        });
+    }
     // If a single pool size was supplied then we are assuming the same sizes across all pools
     if filter_stats.pool_sizes.len() == 1 {
         filter_stats.pool_sizes = vec![filter_stats.pool_sizes[0]; n];
@@ -1313,11 +1335,11 @@ pub fn load_sync<'a, 'b>(
         true => (),
         false => {
             return Err(ImputefError {
-                code: 766,
+                code: 767,
                 message:
                     "Error: the number of pools and the pool sizes do not match in the sync file: "
                         .to_owned()
-                        + &fname,
+                        + fname,
             })
         }
     };

diff --git a/src/vcf.rs b/src/vcf.rs
@@ -626,7 +626,7 @@ pub fn load_vcf<'a, 'b>(
             return Err(ImputefError {
                 code: 823,
                 message: "Error opening the input vcf file: ".to_owned()
-                    + &fname
+                    + fname
                     + " in load_vcf() function.",
             })
         }
@@ -639,7 +639,7 @@ pub fn load_vcf<'a, 'b>(
                 return Err(ImputefError {
                     code: 824,
                     message: "Error reading the input vcf file: ".to_owned()
-                        + &fname
+                        + fname
                         + " in load_vcf() function.",
                 })
             }
@@ -661,6 +661,28 @@ pub fn load_vcf<'a, 'b>(
         }
     }
     let n = pool_names.len();
+    // Check for duplicated pool names
+    let mut unique_pool_names: Vec<String> = vec![];
+    for name_source in pool_names.iter() {
+        let mut duplicated = false;
+        for name_destination in unique_pool_names.iter() {
+            if name_source == name_destination {
+                duplicated = true;
+                break;
+            }
+        }
+        if !duplicated {
+            unique_pool_names.push(name_source.to_string())
+        }
+    }
+    if n > unique_pool_names.len() {
+        return Err(ImputefError {
+            code: 825,
+            message: "Error: there are duplicated pool names in file: ".to_owned()
+                + fname
+                + " in load_vcf() function.",
+        });
+    }
     // If a single pool size was supplied then we are assuming the same sizes across all pools
     if filter_stats.pool_sizes.len() == 1 {
         filter_stats.pool_sizes = vec![filter_stats.pool_sizes[0]; n];
@@ -669,10 +691,10 @@ pub fn load_vcf<'a, 'b>(
         true => (),
         false => {
             return Err(ImputefError {
-                code: 825,
+                code: 826,
                 message: "Error: the number of pools and the pool sizes do not match in file: "
                     .to_owned()
-                    + &fname
+                    + fname
                     + " in load_vcf() function.",
             })
         }
@@ -690,7 +712,7 @@ pub fn load_vcf<'a, 'b>(
         .read_analyse_write(filter_stats, &fname_sync_out, n_threads, vcf_to_sync) {
             Ok(x) => x,
             Err(_) => return Err(ImputefError{
-                code: 826,
+                code: 827,
                 message: "Error converting the vcf into sync via read_analyse_write() method within impute().".to_owned()
             })
         };
@@ -706,9 +728,9 @@ pub fn load_vcf<'a, 'b>(
     .convert_into_genotypes_and_phenotypes(filter_stats, false, n_threads) {
         Ok(x) => x,
         Err(_e) => return Err(ImputefError{
-            code: 827,
+            code: 828,
             message: "Error parsing the input genotype (converted from vcf into sync): ".to_owned() + 
-            &fname +
+            fname +
             " and dummy phenotype data via convert_into_genotypes_and_phenotypes() method within impute()."
         })
     };