| 1 | // ------------------------------------------------------------ |
|---|
| 2 | // |
|---|
| 3 | // Format Conversion Program. |
|---|
| 4 | // |
|---|
| 5 | // Woese Lab., Dept. of Microbiology, UIUC |
|---|
| 6 | // Modified for use in ARB by ARB team |
|---|
| 7 | // |
|---|
| 8 | // ------------------------------------------------------------ |
|---|
| 9 | |
|---|
| 10 | #include "defs.h" |
|---|
| 11 | #include "fun.h" |
|---|
| 12 | #include "global.h" |
|---|
| 13 | |
|---|
| 14 | Convaln_exception *Convaln_exception::thrown = NULL; |
|---|
| 15 | |
|---|
| 16 | struct TypeSwitch { const char *switchtext; Format format; }; |
|---|
| 17 | |
|---|
| 18 | static TypeSwitch convertible_type[] = { // see fconv.cxx@format_spec |
|---|
| 19 | { "GenBank", GENBANK }, |
|---|
| 20 | { "EMBL", EMBL }, |
|---|
| 21 | { "AE2", MACKE }, |
|---|
| 22 | { "SwissProt", SWISSPROT }, |
|---|
| 23 | { "NEXUS", NEXUS }, |
|---|
| 24 | { "PHYLIP", PHYLIP }, |
|---|
| 25 | { "FASTDNAML", FASTDNAML }, |
|---|
| 26 | { "GCG", GCG }, |
|---|
| 27 | { "PRINTABLE", PRINTABLE }, |
|---|
| 28 | }; |
|---|
| 29 | |
|---|
| 30 | static void show_command_line_usage() { |
|---|
| 31 | fputs("Command line usage:\n" |
|---|
| 32 | " $ arb_convert_aln -INFMT input_file -OUTFMT output_file\n" |
|---|
| 33 | " where\n" |
|---|
| 34 | " INFMT may be 'GenBank', 'EMBL', 'AE2' or 'SwissProt' and\n" |
|---|
| 35 | " OUTFMT may be 'GenBank', 'EMBL', 'AE2', 'NEXUS', 'PHYLIP', 'FASTDNAML', 'GCG' or 'Printable'\n" |
|---|
| 36 | " (Note: you may abbreviate the format names)\n" |
|---|
| 37 | " FASTDNAML writes a PHYLIP file with content from STDIN appended at end of first line (used for arb_fastdnaml).\n" |
|---|
| 38 | , stderr); |
|---|
| 39 | } |
|---|
| 40 | |
|---|
| 41 | static void valid_name_or_die(const char *file_name) { |
|---|
| 42 | if (str0len(file_name) <= 0) { |
|---|
| 43 | throw_errorf(152, "illegal file name: %s", file_name); |
|---|
| 44 | } |
|---|
| 45 | } |
|---|
| 46 | static bool file_exists(const char *file_name) { |
|---|
| 47 | FILE *ifp = fopen(file_name, "r"); |
|---|
| 48 | bool exists = ifp != NULL; |
|---|
| 49 | if (ifp) fclose(ifp); |
|---|
| 50 | |
|---|
| 51 | return exists; |
|---|
| 52 | } |
|---|
| 53 | |
|---|
| 54 | static void change_file_suffix(const char *old_file, char *file_name, int type) { |
|---|
| 55 | // Define the default file name by changing suffix. |
|---|
| 56 | int indi, indj; |
|---|
| 57 | |
|---|
| 58 | for (indi = str0len(old_file) - 1; indi >= 0 && old_file[indi] != '.'; indi--) |
|---|
| 59 | if (indi == 0) |
|---|
| 60 | strcpy(file_name, old_file); |
|---|
| 61 | else { |
|---|
| 62 | for (indj = 0; indj < (indi - 1); indj++) |
|---|
| 63 | file_name[indj] = old_file[indj]; |
|---|
| 64 | file_name[indj] = '\0'; |
|---|
| 65 | } |
|---|
| 66 | switch (type) { |
|---|
| 67 | case GENBANK: |
|---|
| 68 | strcat(file_name, ".GB"); |
|---|
| 69 | break; |
|---|
| 70 | case MACKE: |
|---|
| 71 | strcat(file_name, ".aln"); |
|---|
| 72 | break; |
|---|
| 73 | case NEXUS: |
|---|
| 74 | strcat(file_name, ".NEXUS"); |
|---|
| 75 | break; |
|---|
| 76 | case PHYLIP: |
|---|
| 77 | strcat(file_name, ".PHY"); |
|---|
| 78 | break; |
|---|
| 79 | case EMBL: |
|---|
| 80 | strcat(file_name, ".EMBL"); |
|---|
| 81 | break; |
|---|
| 82 | case PRINTABLE: |
|---|
| 83 | strcat(file_name, ".prt"); |
|---|
| 84 | break; |
|---|
| 85 | default: |
|---|
| 86 | strcat(file_name, ".???"); |
|---|
| 87 | } |
|---|
| 88 | } |
|---|
| 89 | |
|---|
| 90 | static void ask_for_conversion_params(FormattedFile& in, FormattedFile& out) { |
|---|
| 91 | char temp[LINESIZE]; |
|---|
| 92 | char choice[LINESIZE]; |
|---|
| 93 | |
|---|
| 94 | fputs("---------------------------------------------------------------\n" |
|---|
| 95 | "\n" |
|---|
| 96 | " convert_aln - an alignment and file converter written by\n" |
|---|
| 97 | " WenMin Kuan for the RDP database project.\n" |
|---|
| 98 | "\n" |
|---|
| 99 | " Modified for use in ARB by Oliver Strunk & Ralf Westram\n" |
|---|
| 100 | " Report errors or deficiencies to devel@arb-home.de\n" |
|---|
| 101 | "\n" |
|---|
| 102 | , stderr); |
|---|
| 103 | show_command_line_usage(); |
|---|
| 104 | fputs("\n" |
|---|
| 105 | "---------------------------------------------------------------\n" |
|---|
| 106 | "\n" |
|---|
| 107 | "Select input format (<CR> means default)\n" |
|---|
| 108 | "\n" |
|---|
| 109 | " (1) GenBank [default]\n" |
|---|
| 110 | " (2) EMBL\n" |
|---|
| 111 | " (3) AE2\n" |
|---|
| 112 | " (4) SwissProt\n" |
|---|
| 113 | " (5) Quit\n" |
|---|
| 114 | " ? " |
|---|
| 115 | , stderr); |
|---|
| 116 | |
|---|
| 117 | Getstr(choice, LINESIZE); |
|---|
| 118 | { |
|---|
| 119 | Format inType = UNKNOWN; |
|---|
| 120 | switch (choice[0]) { |
|---|
| 121 | case '\0': // [default] |
|---|
| 122 | case '1': inType = GENBANK; break; |
|---|
| 123 | case '2': inType = EMBL; break; |
|---|
| 124 | case '3': inType = MACKE; break; |
|---|
| 125 | case '4': inType = SWISSPROT; break; |
|---|
| 126 | case '5': exit(0); // ok - interactive mode only |
|---|
| 127 | default: throw_errorf(16, "Unknown input format selection '%s'", choice); |
|---|
| 128 | } |
|---|
| 129 | |
|---|
| 130 | fputs("\nInput file name? ", stderr); |
|---|
| 131 | Getstr(temp, LINESIZE); |
|---|
| 132 | in.init(temp, inType); |
|---|
| 133 | } |
|---|
| 134 | |
|---|
| 135 | valid_name_or_die(temp); |
|---|
| 136 | if (!file_exists(temp)) throw_error(77, "Input file not found"); |
|---|
| 137 | |
|---|
| 138 | // output file information |
|---|
| 139 | fputs("\n" |
|---|
| 140 | "Select output format (<CR> means default)\n" |
|---|
| 141 | "\n" |
|---|
| 142 | " (1) GenBank\n" |
|---|
| 143 | " (2) EMBL\n" |
|---|
| 144 | " (3) AE2 [default]\n" |
|---|
| 145 | " (4) NEXUS (Paup)\n" |
|---|
| 146 | " (5) PHYLIP\n" |
|---|
| 147 | " (6) GCG\n" |
|---|
| 148 | " (7) Printable\n" |
|---|
| 149 | " (8) Quit\n" |
|---|
| 150 | " ? ", stderr); |
|---|
| 151 | |
|---|
| 152 | Getstr(choice, LINESIZE); |
|---|
| 153 | { |
|---|
| 154 | Format ouType = UNKNOWN; |
|---|
| 155 | switch (choice[0]) { |
|---|
| 156 | case '1': ouType = GENBANK; break; |
|---|
| 157 | case '2': ouType = EMBL; break; |
|---|
| 158 | case '\0': // [default] |
|---|
| 159 | case '3': ouType = MACKE; break; |
|---|
| 160 | case '4': ouType = NEXUS; break; |
|---|
| 161 | case '5': ouType = PHYLIP; break; |
|---|
| 162 | case '6': ouType = GCG; break; |
|---|
| 163 | case '7': ouType = PRINTABLE; break; |
|---|
| 164 | case '8': exit(0); // ok - interactive mode only |
|---|
| 165 | default: throw_errorf(66, "Unknown output format selection '%s'", choice); |
|---|
| 166 | } |
|---|
| 167 | change_file_suffix(in.name(), temp, ouType); |
|---|
| 168 | if (ouType != GCG) { |
|---|
| 169 | fprintf(stderr, "\nOutput file name [%s]? ", temp); |
|---|
| 170 | Getstr(temp, LINESIZE); |
|---|
| 171 | if (str0len(temp) == 0) |
|---|
| 172 | change_file_suffix(in.name(), temp, ouType); |
|---|
| 173 | } |
|---|
| 174 | out.init(temp, ouType); |
|---|
| 175 | } |
|---|
| 176 | } |
|---|
| 177 | |
|---|
| 178 | static int strcasecmp_start(const char *s1, const char *s2) { |
|---|
| 179 | int cmp = 0; |
|---|
| 180 | for (int p = 0; !cmp; p++) { |
|---|
| 181 | cmp = tolower(s1[p])-tolower(s2[p]); |
|---|
| 182 | if (!s1[p]) return 0; |
|---|
| 183 | } |
|---|
| 184 | return cmp; |
|---|
| 185 | } |
|---|
| 186 | |
|---|
| 187 | static bool is_abbrev_switch(const char *arg, const char *switchtext) { |
|---|
| 188 | return arg[0] == '-' && strcasecmp_start(arg+1, switchtext) == 0; |
|---|
| 189 | } |
|---|
| 190 | |
|---|
| 191 | static Format parse_type(const char *arg) { |
|---|
| 192 | for (size_t i = 0; i<ARRAY_ELEMS(convertible_type); ++i) { |
|---|
| 193 | const TypeSwitch& type = convertible_type[i]; |
|---|
| 194 | if (is_abbrev_switch(arg, type.switchtext)) { |
|---|
| 195 | return type.format; |
|---|
| 196 | } |
|---|
| 197 | } |
|---|
| 198 | return UNKNOWN; |
|---|
| 199 | } |
|---|
| 200 | |
|---|
| 201 | static Format parse_intype(const char *arg) { |
|---|
| 202 | Format type = parse_type(arg); |
|---|
| 203 | if (!is_input_format(type)) throw_errorf(65, "Unsupported input file type '%s'", arg); |
|---|
| 204 | if (type == UNKNOWN) throw_errorf(67, "UNKNOWN input file type '%s'", arg); |
|---|
| 205 | return type; |
|---|
| 206 | } |
|---|
| 207 | |
|---|
| 208 | static Format parse_outtype(const char *arg) { |
|---|
| 209 | Format type = parse_type(arg); |
|---|
| 210 | if (type == UNKNOWN) throw_errorf(68, "UNKNOWN output file type '%s'", arg); |
|---|
| 211 | return type; |
|---|
| 212 | } |
|---|
| 213 | |
|---|
| 214 | static bool is_help_req(const char *arg) { |
|---|
| 215 | return strcmp(arg, "-h") == 0 || strcmp(arg, "--help") == 0; |
|---|
| 216 | } |
|---|
| 217 | static bool command_line_conversion(int argc, const char * const *argv, FormattedFile& in, FormattedFile& out) { |
|---|
| 218 | for (int c = 1; c<argc; c++) { |
|---|
| 219 | if (is_help_req(argv[c])) { |
|---|
| 220 | show_command_line_usage(); |
|---|
| 221 | return false; |
|---|
| 222 | } |
|---|
| 223 | } |
|---|
| 224 | |
|---|
| 225 | if (argc != 5) throw_errorf(69, "arb_convert_aln expects exactly 4 parameters (you specified %i). Try '--help'", argc-1); |
|---|
| 226 | |
|---|
| 227 | in.init(argv[2], parse_intype(argv[1])); |
|---|
| 228 | out.init(argv[4], parse_outtype(argv[3])); |
|---|
| 229 | |
|---|
| 230 | return true; |
|---|
| 231 | } |
|---|
| 232 | |
|---|
| 233 | static void do_conversion(const FormattedFile& in, const FormattedFile& out) { |
|---|
| 234 | #ifdef CALOG |
|---|
| 235 | fprintf(stderr, "\n\nConvert file %s to file %s.\n", in.name(), out.name()); |
|---|
| 236 | #endif |
|---|
| 237 | |
|---|
| 238 | // check if output file exists and filename's validation |
|---|
| 239 | valid_name_or_die(out.name()); |
|---|
| 240 | if (file_exists(out.name())) warningf(151, "Output file %s exists, will be overwritten.", out.name()); |
|---|
| 241 | |
|---|
| 242 | // file format transfer... |
|---|
| 243 | convert(in, out); |
|---|
| 244 | } |
|---|
| 245 | |
|---|
| 246 | int ARB_main(int argc, const char *argv[]) { |
|---|
| 247 | int exitcode = EXIT_SUCCESS; |
|---|
| 248 | try { |
|---|
| 249 | FormattedFile in; |
|---|
| 250 | FormattedFile out; |
|---|
| 251 | |
|---|
| 252 | if (argc < 2) { |
|---|
| 253 | ask_for_conversion_params(in, out); |
|---|
| 254 | do_conversion(in, out); |
|---|
| 255 | } |
|---|
| 256 | else { |
|---|
| 257 | if (command_line_conversion(argc, argv, in, out)) { |
|---|
| 258 | do_conversion(in, out); |
|---|
| 259 | } |
|---|
| 260 | } |
|---|
| 261 | } |
|---|
| 262 | catch (Convaln_exception& err) { |
|---|
| 263 | fprintf(stderr, "ERROR(%d): %s\n", err.get_code(), err.get_msg()); |
|---|
| 264 | exitcode = EXIT_FAILURE; |
|---|
| 265 | } |
|---|
| 266 | return exitcode; |
|---|
| 267 | } |
|---|
| 268 | |
|---|
| 269 | // -------------------------------------------------------------------------------- |
|---|
| 270 | |
|---|
| 271 | #ifdef UNIT_TESTS |
|---|
| 272 | #include <test_unit.h> |
|---|
| 273 | |
|---|
| 274 | void TEST_BASIC_switch_parsing() { |
|---|
| 275 | TEST_ASSERT(strcasecmp_start("GenBank", "GenBank") == 0); |
|---|
| 276 | TEST_ASSERT(strcasecmp_start("GEnbaNK", "genBANK") == 0); |
|---|
| 277 | TEST_ASSERT(strcasecmp_start("Ge", "GenBank") == 0); |
|---|
| 278 | TEST_ASSERT(strcasecmp_start("GenBank", "NEXUS") < 0); |
|---|
| 279 | TEST_ASSERT(strcasecmp_start("NEXUS", "GenBank") > 0); |
|---|
| 280 | |
|---|
| 281 | TEST_ASSERT(!is_abbrev_switch("notAswitch", "notAswitch")); |
|---|
| 282 | TEST_ASSERT(!is_abbrev_switch("-GenbankPlus", "Genbank")); |
|---|
| 283 | TEST_ASSERT(!is_abbrev_switch("-Ge", "NEXUS")); |
|---|
| 284 | |
|---|
| 285 | TEST_ASSERT(is_abbrev_switch("-Ge", "Genbank")); |
|---|
| 286 | TEST_ASSERT(is_abbrev_switch("-N", "NEXUS")); |
|---|
| 287 | TEST_ASSERT(is_abbrev_switch("-NEXUS", "NEXUS")); |
|---|
| 288 | |
|---|
| 289 | TEST_ASSERT_EQUAL(parse_outtype("-PH"), PHYLIP); |
|---|
| 290 | TEST_ASSERT_EQUAL(parse_outtype("-PHYLIP"), PHYLIP); |
|---|
| 291 | TEST_ASSERT_EQUAL(parse_outtype("-phylip"), PHYLIP); |
|---|
| 292 | } |
|---|
| 293 | |
|---|
| 294 | #endif // UNIT_TESTS |
|---|
