22 files changed, 181 insertions, 71 deletions

diff --git a/src/load/load_influenza_aa_dat.c b/src/load/load_influenza_aa_dat.c
new file mode 100644
index 0000000..8bf47aa
--- a/dev/null
+++ b/src/load/load_influenza_aa_dat.c
@@ -0,0 +1,254 @@
+/*
+ * Load the influnza_aa.dat tab delimited text file into an HDF5
+ * binary table.
+ *
+ * todo: Handle NULL values occuring in numeric fields.
+ */
+
+#include "load_influenza_aa_dat.h"
+#include "error/check_error.h"
+#include "error/check_h5_error.h"
+#include <hdf5_hl.h>
+#include <string.h>
+#include <stdlib.h>
+
+#define NFIELDS (hsize_t) 11
+#define TABLE_NAME "influenza_aa.dat"
+
+void
+load_influenza_aa_dat (hid_t file_id)
+{
+  /*
+   * Model the data using native types.
+   */
+  typedef struct
+  {
+    char genbank_accession_number[9];
+    char host[15];
+    int genome_segment_number;
+    char subtype[7];
+    char country[25];
+    int year;
+    int sequence_length;
+    char virus_name[196];
+    char age[17];
+    char gender[6];
+    char full_length_indicator[4];
+  } supplementary_data;
+
+  /*
+   * Use an HDF5 Table for storage.
+   * http://www.hdfgroup.org/HDF5/Tutor/h5table.html
+   */
+
+  /*
+   * "Calculate the size and the offsets of our struct members in
+   * memory."
+   */
+  size_t dst_size = sizeof (supplementary_data);
+  size_t dst_offset[NFIELDS] =
+    { HOFFSET (supplementary_data, genbank_accession_number),
+    HOFFSET (supplementary_data, host),
+    HOFFSET (supplementary_data, genome_segment_number),
+    HOFFSET (supplementary_data, subtype),
+    HOFFSET (supplementary_data, country),
+    HOFFSET (supplementary_data, year),
+    HOFFSET (supplementary_data, sequence_length),
+    HOFFSET (supplementary_data, virus_name),
+    HOFFSET (supplementary_data, age),
+    HOFFSET (supplementary_data, gender),
+    HOFFSET (supplementary_data, full_length_indicator)
+  };
+
+  supplementary_data dst_buf[1];
+
+  size_t dst_sizes[NFIELDS] = { sizeof (dst_buf[0].genbank_accession_number),
+    sizeof (dst_buf[0].host),
+    sizeof (dst_buf[0].genome_segment_number),
+    sizeof (dst_buf[0].subtype),
+    sizeof (dst_buf[0].country),
+    sizeof (dst_buf[0].year),
+    sizeof (dst_buf[0].sequence_length),
+    sizeof (dst_buf[0].virus_name),
+    sizeof (dst_buf[0].age),
+    sizeof (dst_buf[0].gender),
+    sizeof (dst_buf[0].full_length_indicator)
+  };
+
+  /*
+   * Map the native types to HDF5 types for each field.
+   */
+  hid_t field_type[NFIELDS];
+
+  hid_t genbank_accession_number_type = H5Tcopy (H5T_C_S1);
+  H5Tset_size (genbank_accession_number_type, 9);
+  field_type[0] = genbank_accession_number_type;
+
+  hid_t host_type = H5Tcopy (H5T_C_S1);
+  H5Tset_size (host_type, 15);
+  field_type[1] = host_type;
+
+  field_type[2] = H5T_NATIVE_INT;
+
+  hid_t subtype_type = H5Tcopy (H5T_C_S1);
+  H5Tset_size (subtype_type, 7);
+  field_type[3] = subtype_type;
+
+  hid_t country_type = H5Tcopy (H5T_C_S1);
+  H5Tset_size (country_type, 25);
+  field_type[4] = country_type;
+
+  field_type[5] = H5T_NATIVE_INT;
+
+  field_type[6] = H5T_NATIVE_INT;
+
+  hid_t virus_name_type = H5Tcopy (H5T_C_S1);
+  H5Tset_size (virus_name_type, 196);
+  field_type[7] = virus_name_type;
+
+  hid_t age_type = H5Tcopy (H5T_C_S1);
+  H5Tset_size (age_type, 17);
+  field_type[8] = age_type;
+
+  hid_t gender_type = H5Tcopy (H5T_C_S1);
+  H5Tset_size (gender_type, 6);
+  field_type[9] = gender_type;
+
+  hid_t full_length_indicator_type = H5Tcopy (H5T_C_S1);
+  H5Tset_size (full_length_indicator_type, 4);
+  field_type[10] = full_length_indicator_type;
+
+  /*
+   * Labels used for the fields in the table.
+   */
+  const char *field_names[NFIELDS] = { "GenBank accession number",
+    "Host",
+    "Genome segment number",
+    "Subtype",
+    "Country",
+    "Year",
+    "Sequence length",
+    "Virus name",
+    "Age",
+    "Gender",
+    "Full-length Indicator"
+  };
+
+  /*
+   * Table storage options.
+   */
+  hsize_t chunk_size = 10;
+  int *fill_data = NULL;
+  int compress = 0;
+
+  /*
+   * Insert the records.
+   */
+  supplementary_data p_data;
+  FILE *dat = fopen ("/home/don/exp004/genomes/INFLUENZA/influenza_aa.dat",
+                     "r");
+  if (dat == NULL)
+    check_error (__FILE__, __LINE__);
+  char *line = NULL;
+  size_t len = 0;
+  int current_line = 0;
+
+  while (getline (&line, &len, dat) != -1)
+    {
+
+      current_line++;
+      char *running = strdup (line);
+      char *token = NULL;
+
+      /*
+       * Parse the line, handling the case of empty fields represented
+       * by sequential delimiters.
+       */
+      strncpy (p_data.genbank_accession_number, strsep (&running, "\t"),
+               sizeof (p_data.genbank_accession_number));
+
+      strncpy (p_data.host, strsep (&running, "\t"), sizeof (p_data.host));
+
+      token = strsep (&running, "\t");
+      if (strcmp (token, "\0") == 0)
+        p_data.genome_segment_number = 0;
+      else
+        p_data.genome_segment_number = atoi (token);
+
+      strncpy (p_data.subtype, strsep (&running, "\t"),
+               sizeof (p_data.subtype));
+
+      strncpy (p_data.country, strsep (&running, "\t"),
+               sizeof (p_data.country));
+
+      /*
+       * Convert the year field from text to numeric.  Unknown and empty
+       * values are assigned a numeric value of zero.
+       */
+      token = strsep (&running, "\t");
+      if (strcmp (token, "\0") == 0)
+        p_data.year = 0;
+      else if (strcmp (token, "unknown") == 0)
+        p_data.year = 0;
+      else if (strcmp (token, "NON") == 0)
+        p_data.year = 0;
+      else
+        p_data.year = atoi (token);
+
+      token = strsep (&running, "\t");
+      if (strcmp (token, "\0") == 0)
+        p_data.sequence_length = 0;
+      else
+        p_data.sequence_length = atoi (token);
+
+      strncpy (p_data.virus_name, strsep (&running, "\t"),
+               sizeof (p_data.virus_name));
+
+      strncpy (p_data.age, strsep (&running, "\t"), sizeof (p_data.age));
+
+      strncpy (p_data.gender, strsep (&running, "\t"),
+               sizeof (p_data.gender));
+
+      strncpy (p_data.full_length_indicator, strsep (&running, "\t"),
+               sizeof (p_data.full_length_indicator));
+
+      if (current_line == 1)
+        {
+          herr_t status = H5TBmake_table ("influenza_aa.dat", file_id,
+                                          TABLE_NAME, NFIELDS, 1, dst_size,
+                                          field_names, dst_offset, field_type,
+                                          chunk_size, fill_data, compress,
+                                          &p_data);
+          if (status < 0)
+            check_h5_error (status, __FILE__, __LINE__);
+        }
+      else
+        {
+          herr_t status =
+            H5TBappend_records (file_id, TABLE_NAME, 1, dst_size,
+                                dst_offset, dst_sizes, &p_data);
+          if (status < 0)
+            check_h5_error (status, __FILE__, __LINE__);
+        }
+
+      if (running)
+        free (running);
+
+    }
+
+  if (line)
+    free (line);
+
+  fclose (dat);
+
+  H5Tclose (genbank_accession_number_type);
+  H5Tclose (host_type);
+  H5Tclose (subtype_type);
+  H5Tclose (country_type);
+  H5Tclose (virus_name_type);
+  H5Tclose (age_type);
+  H5Tclose (gender_type);
+  H5Tclose (full_length_indicator_type);
+
+  return;
+}