1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/ameriflux/02.read_ascii.ncl Mon Jan 26 22:08:20 2009 -0500
1.3 @@ -0,0 +1,377 @@
1.4 +;----------------------------------------------------------------------
1.5 +; add year to output
1.6 +;
1.7 +; This example reads an ASCII file that is formatted a specific way, and
1.8 +; writes out the results to a netCDF file.
1.9 +;
1.10 +; The first line in the ASCII file must be a header, with each field
1.11 +; separated by a single character delimiter (like a ","). The rest of
1.12 +; the file must be such that each row contains all fields, each
1.13 +; separated by the designated delimiter.
1.14 +;
1.15 +; The fields can be integer, float, double, character, or string.
1.16 +; String fields cannot be written to a netCDF file. They have to
1.17 +; be read in as character arrays and written out that way.
1.18 +;----------------------------------------------------------------------
1.19 +
1.20 +;----------------------------------------------------------------------
1.21 +; This function returns the index locations of the given delimiter
1.22 +; in a row or several rows of strings.
1.23 +;----------------------------------------------------------------------
1.24 +function delim_indices(strings,nfields,delimiter)
1.25 +local cstrings, cdelim
1.26 +begin
1.27 + nrows = dimsizes(strings)
1.28 +;
1.29 +; Handle special case if we only have one string. Make sure it
1.30 +; is put into a 2D array.
1.31 +;
1.32 + if(nrows.eq.1) then
1.33 + cstrings = new((/1,strlen(strings)+1/),character)
1.34 + end if
1.35 +
1.36 + cstrings = stringtochar(strings) ; Convert to characters.
1.37 + cdelim = stringtochar(delimiter) ; Convert delimiter to character.
1.38 +;
1.39 +; Som error checking here. Make sure delimiter is one character.
1.40 +;
1.41 + nc = dimsizes(cdelim)
1.42 + rank = dimsizes(nc)
1.43 + if(rank.ne.1.or.(rank.eq.1.and.nc.ne.2)) then
1.44 + print("delim_indices: fatal: the delimiter you've selected")
1.45 + print("must be a single character. Can't continue.")
1.46 + exit
1.47 + end if
1.48 +
1.49 +;
1.50 +; Create array to hold indices of delimiter locations, and then loop
1.51 +; through each row and find all the delimiters. Make sure each row has
1.52 +; the correct number of delimiters.
1.53 +;
1.54 + ndelims = nfields-1
1.55 + cindices = new((/nrows,ndelims/),integer)
1.56 + do i = 0, nrows-1
1.57 + ii = ind(cstrings(i,:).eq.cdelim(0))
1.58 +;
1.59 +; Make sure there were delimiters on this row. If not, we just quit.
1.60 +; This could probably be modified to do this more gracefully.
1.61 +;
1.62 + if(any(ismissing(ii))) then
1.63 + print("delim_indices: fatal: I didn't find any delimiters")
1.64 + print("('" + delimiter + "') on row " + i + ". Can't continue.")
1.65 + exit
1.66 + end if
1.67 + if(dimsizes(ii).ne.ndelims) then
1.68 + print("delim_indices: fatal: I expected to find " + ndelims)
1.69 + print("delimiters on row " + i + ". Instead, I found " + dimsizes(ii) + ".")
1.70 + print("Can't continue.")
1.71 + exit
1.72 + end if
1.73 +
1.74 + cindices(i,:) = ii
1.75 +
1.76 + delete(ii) ; For next time through loop
1.77 + end do
1.78 +
1.79 + return(cindices)
1.80 +end
1.81 +
1.82 +;----------------------------------------------------------------------
1.83 +; This function reads in a particular field from a string array,
1.84 +; given the field number to read (fields start at #1 and go to #nfield),
1.85 +; and the indices of the delimiters.
1.86 +;
1.87 +; It returns either an integer, float, double, character, or a string,
1.88 +; depending on the input flag "return_type".
1.89 +;----------------------------------------------------------------------
1.90 +function read_field(strings,ifield,indices,return_type)
1.91 +local nstring, cstrings, nf, tmp_str
1.92 +begin
1.93 + nrows = dimsizes(strings)
1.94 +;
1.95 +; Handle special case if we only have one string. Make sure it
1.96 +; is put into a 2D array.
1.97 +;
1.98 + if(nrows.eq.1) then
1.99 + cstrings = new((/1,strlen(strings)+1/),character)
1.100 + end if
1.101 +
1.102 + cstrings = stringtochar(strings)
1.103 + nf = dimsizes(indices(0,:))+1 ; indices is nrows x (nfields-1)
1.104 +
1.105 +;
1.106 +; Error checking. Make sure user has entered a valid field.
1.107 +;
1.108 + if(ifield.le.0.or.ifield.gt.nf) then
1.109 + print("read_field: fatal: you've selected a field that is")
1.110 + print("out-of-range of the number of fields that you have (" + nf + ").")
1.111 + exit
1.112 + end if
1.113 +
1.114 +;
1.115 +; Set up array to return. For string, int, float, or double arrays,
1.116 +; we don't have to do anything special. For character arrays,
1.117 +; however, we do.
1.118 +;
1.119 + if(return_type.ne."character") then
1.120 + return_array = new(nrows,return_type)
1.121 + else
1.122 +;
1.123 +; We don't know what the biggest character array is at this point, so
1.124 +; make it bigger than necessary, and then resize later as necessary.
1.125 +;
1.126 + tmp_return_array = new((/nrows,dimsizes(cstrings(0,:))/),"character")
1.127 +
1.128 + max_len = 0 ; Use to keep track of max lengths of strings.
1.129 + end if
1.130 +
1.131 + do i = 0,nrows-1
1.132 +;
1.133 +; Special case of first field in row.
1.134 +;
1.135 + if(ifield.eq.1) then
1.136 + ibeg = 0
1.137 + iend = indices(i,ifield-1)-1
1.138 + else
1.139 +;
1.140 +; Special case of first field in row.
1.141 +;
1.142 + if(ifield.eq.nf) then
1.143 + ibeg = indices(i,ifield-2)+1
1.144 + iend = dimsizes(cstrings(i,:))-1
1.145 +;
1.146 +; Any field between first and last field.
1.147 +;
1.148 + else
1.149 + ibeg = indices(i,ifield-2)+1
1.150 + iend = indices(i,ifield-1)-1
1.151 + end if
1.152 + end if
1.153 +;
1.154 +; Here's the code that pulls off the correct string, and converts it
1.155 +; to float if desired.
1.156 +;
1.157 + if(return_type.eq."integer") then
1.158 + return_array(i) = stringtointeger(chartostring(cstrings(i,ibeg:iend)))
1.159 + end if
1.160 + if(return_type.eq."float") then
1.161 + return_array(i) = stringtofloat(chartostring(cstrings(i,ibeg:iend)))
1.162 + end if
1.163 + if(return_type.eq."double") then
1.164 + return_array(i) = stringtodouble(chartostring(cstrings(i,ibeg:iend)))
1.165 + end if
1.166 + if(return_type.eq."string") then
1.167 + return_array(i) = chartostring(cstrings(i,ibeg:iend))
1.168 + end if
1.169 + if(return_type.eq."character") then
1.170 + if( (iend-ibeg+1) .gt. max_len) then
1.171 + max_len = iend-ibeg+1
1.172 + end if
1.173 + tmp_return_array(i,0:iend-ibeg) = cstrings(i,ibeg:iend)
1.174 + end if
1.175 + end do
1.176 +
1.177 + if(return_type.eq."character") then
1.178 + return_array = new((/nrows,max_len/),"character")
1.179 + return_array = tmp_return_array(:,0:max_len-1)
1.180 + end if
1.181 +
1.182 + return(return_array)
1.183 +end
1.184 +
1.185 +
1.186 +;----------------------------------------------------------------------
1.187 +; This function reads in string fields only to get the maximum string
1.188 +; length.
1.189 +;----------------------------------------------------------------------
1.190 +function get_maxlen(strings,ifield,indices)
1.191 +local nstring, cstrings, nf, tmp_str
1.192 +begin
1.193 + nrows = dimsizes(strings)
1.194 +;
1.195 +; Handle special case if we only have one string. Make sure it
1.196 +; is put into a 2D array.
1.197 +;
1.198 + if(nrows.eq.1) then
1.199 + cstrings = new((/1,strlen(strings)+1/),character)
1.200 + end if
1.201 +
1.202 + cstrings = stringtochar(strings)
1.203 + nf = dimsizes(indices(0,:))+1 ; indices is nrows x (nfields-1)
1.204 +
1.205 +;
1.206 +; Error checking. Make sure user has entered a valid field.
1.207 +;
1.208 + if(ifield.le.0.or.ifield.gt.nf) then
1.209 + print("read_field: fatal: you've selected a field that is")
1.210 + print("out-of-range of the number of fields that you have (" + nf + ").")
1.211 + exit
1.212 + end if
1.213 +;
1.214 +; We don't know what the biggest character array is at this point, so
1.215 +; make it bigger than necessary, and then resize later as necessary.
1.216 +;
1.217 + tmp_return_array = new((/nrows,dimsizes(cstrings(0,:))/),"character")
1.218 +
1.219 + max_len = 0 ; Use to keep track of max lengths of strings.
1.220 +
1.221 + do i = 0,nrows-1
1.222 +;
1.223 +; Special case of first field in row.
1.224 +;
1.225 + if(ifield.eq.1) then
1.226 + ibeg = 0
1.227 + iend = indices(i,ifield-1)-1
1.228 + else
1.229 +;
1.230 +; Special case of first field in row.
1.231 +;
1.232 + if(ifield.eq.nf) then
1.233 + ibeg = indices(i,ifield-2)+1
1.234 + iend = dimsizes(cstrings(i,:))-1
1.235 +;
1.236 +; Any field between first and last field.
1.237 +;
1.238 + else
1.239 + ibeg = indices(i,ifield-2)+1
1.240 + iend = indices(i,ifield-1)-1
1.241 + end if
1.242 + end if
1.243 + if( (iend-ibeg+1) .gt. max_len) then
1.244 + max_len = iend-ibeg+1
1.245 + end if
1.246 + end do
1.247 +
1.248 + return(max_len)
1.249 +end
1.250 +
1.251 +;----------------------------------------------------------------------
1.252 +; Main code.
1.253 +;----------------------------------------------------------------------
1.254 +begin
1.255 +
1.256 +; Set up defaults here. We are hard-coding the field types here.
1.257 +; You can set up this script to try to determine the field types
1.258 +; automatically, but this is a bit tedious. Maybe later.
1.259 +
1.260 + station = "USHa1"
1.261 + year = 1991
1.262 + nfields = 30 ; # of fields
1.263 + delimiter = "," ; field delimiter
1.264 +
1.265 + filename = station+year+"_L4_m.txt" ; ASCII" file to read.
1.266 + cdf_file = station+year+"_L4_m.nc" ; netCDF file to write.
1.267 +
1.268 +; In this case, fields #2-#2 are integers,
1.269 +; and the rest of the fields are floats.
1.270 +
1.271 + var_types = new(nfields,string)
1.272 + var_strlens = new(nfields,integer) ; var to hold strlens, just in case.
1.273 +
1.274 + var_types = "float" ; Most are floats.
1.275 + var_types(0:1) = "integer"
1.276 +
1.277 + if(isfilepresent(cdf_file))
1.278 + print("Warning: '" + cdf_file + "' exists. Will remove it.")
1.279 + system("/bin/rm " + cdf_file)
1.280 + end if
1.281 +
1.282 +; Read in data as strings. This will create a string array that has the
1.283 +; same number of strings as there are rows in the file. We will then need
1.284 +; to parse each string later.
1.285 +
1.286 + read_data = asciiread(filename,-1,"string")
1.287 + header = read_data(0) ; Header. Use for variable names.
1.288 + data = read_data(1:) ; Get rid of first line which is a header.
1.289 + nrows = dimsizes(data) ; Number of rows.
1.290 +
1.291 +; Read in locations of delimiters in each string row.
1.292 +
1.293 + hindices = delim_indices(header,nfields,delimiter) ; header row
1.294 + dindices = delim_indices(data,nfields,delimiter) ; rest of file
1.295 +
1.296 + print (hindices)
1.297 + print (dindices)
1.298 +
1.299 +; Read in the field names which will become variable names on
1.300 +; the netCDF file.
1.301 +
1.302 + var_names = new(nfields,string)
1.303 +
1.304 + do i=0,nfields-1
1.305 + var_names(i) = read_field(header,i+1,hindices,"string")
1.306 + end do
1.307 +
1.308 +; Write out this netCDF file efficiently so it will be faster.
1.309 +; Try to predefine everything before you write to it.
1.310 +
1.311 + f = addfile(cdf_file,"c")
1.312 + setfileoption(f,"DefineMode",True) ; Enter predefine phase.
1.313 +
1.314 +; Write global attributes to file. It's okay to do this before
1.315 +; predefining the file's variables. We are still in "define" mode.
1.316 +
1.317 + fAtt = True
1.318 + fAtt@description = "Data read in from " + filename + " ASCII file."
1.319 + fAtt@creation_date = systemfunc ("date")
1.320 + fileattdef( f, fAtt )
1.321 +
1.322 +; Write dimension names to file. If there are no character variables,
1.323 +; then there's only one dimension name ("nvalues").
1.324 +
1.325 + nyear = -1
1.326 + nmonth = 12
1.327 +
1.328 + dim_names = (/ "year", "month" /)
1.329 + dim_sizes = (/ nyear , nmonth /)
1.330 + dimUnlim = (/ True , False /)
1.331 +
1.332 + filedimdef( f, dim_names, dim_sizes, dimUnlim )
1.333 +
1.334 +; Define each variable on the file.
1.335 +
1.336 + filevardef( f, "year", "integer", "year" )
1.337 +
1.338 +; Don't deal with variable Month (i=0).
1.339 +
1.340 + do i=1,nfields-1
1.341 + filevardef(f, var_names(i), var_types(i), dim_names)
1.342 + end do
1.343 +
1.344 +; Loop through each field, read the values for that field, print
1.345 +; information about the variable, and then write it to the netCDF
1.346 +; file.
1.347 +
1.348 + do i=1,nfields-1
1.349 + ifield = i+1 ; Fields start at #1, not #0.
1.350 +
1.351 +; Note: you can't write strings to a netCDF file, so these have
1.352 +; to be written out as character arrays.
1.353 +
1.354 + tmp_data = new((/1,nmonth/),var_types(i))
1.355 +
1.356 + out_data = read_field(data,ifield,dindices,var_types(i))
1.357 +
1.358 + tmp_data(0,:) = out_data(:)
1.359 +
1.360 +; Print some info about the variable.
1.361 +
1.362 + print("")
1.363 + print("Writing variable '" + var_names(i) + "' (field #" + ifield + ").")
1.364 + print("Type is " + var_types(i) + ".")
1.365 + print("min/max = " + min(tmp_data) + "/" + max(tmp_data))
1.366 +
1.367 + if(any(ismissing(tmp_data))) then
1.368 + print("This variable does contain missing values.")
1.369 + else
1.370 + print("This variable doesn't contain missing values.")
1.371 + end if
1.372 +
1.373 + f->$var_names(i)$ = tmp_data ; Write to netCDF file.
1.374 +
1.375 + delete(tmp_data) ; Delete for next round.
1.376 + delete(out_data)
1.377 + end do
1.378 +
1.379 + f->year = year
1.380 +end