1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/ameriflux/03.read_ascii.ncl Mon Jan 26 22:08:20 2009 -0500
1.3 @@ -0,0 +1,387 @@
1.4 +;----------------------------------------------------------------------
1.5 +; add Month to output
1.6 +; add lat, lon to output
1.7 +; add year to output
1.8 +;
1.9 +; This example reads an ASCII file that is formatted a specific way, and
1.10 +; writes out the results to a netCDF file.
1.11 +;
1.12 +; The first line in the ASCII file must be a header, with each field
1.13 +; separated by a single character delimiter (like a ","). The rest of
1.14 +; the file must be such that each row contains all fields, each
1.15 +; separated by the designated delimiter.
1.16 +;
1.17 +; The fields can be integer, float, double, character, or string.
1.18 +; String fields cannot be written to a netCDF file. They have to
1.19 +; be read in as character arrays and written out that way.
1.20 +;----------------------------------------------------------------------
1.21 +
1.22 +;----------------------------------------------------------------------
1.23 +; This function returns the index locations of the given delimiter
1.24 +; in a row or several rows of strings.
1.25 +;----------------------------------------------------------------------
1.26 +function delim_indices(strings,nfields,delimiter)
1.27 +local cstrings, cdelim
1.28 +begin
1.29 + nrows = dimsizes(strings)
1.30 +;
1.31 +; Handle special case if we only have one string. Make sure it
1.32 +; is put into a 2D array.
1.33 +;
1.34 + if(nrows.eq.1) then
1.35 + cstrings = new((/1,strlen(strings)+1/),character)
1.36 + end if
1.37 +
1.38 + cstrings = stringtochar(strings) ; Convert to characters.
1.39 + cdelim = stringtochar(delimiter) ; Convert delimiter to character.
1.40 +;
1.41 +; Som error checking here. Make sure delimiter is one character.
1.42 +;
1.43 + nc = dimsizes(cdelim)
1.44 + rank = dimsizes(nc)
1.45 + if(rank.ne.1.or.(rank.eq.1.and.nc.ne.2)) then
1.46 + print("delim_indices: fatal: the delimiter you've selected")
1.47 + print("must be a single character. Can't continue.")
1.48 + exit
1.49 + end if
1.50 +
1.51 +;
1.52 +; Create array to hold indices of delimiter locations, and then loop
1.53 +; through each row and find all the delimiters. Make sure each row has
1.54 +; the correct number of delimiters.
1.55 +;
1.56 + ndelims = nfields-1
1.57 + cindices = new((/nrows,ndelims/),integer)
1.58 + do i = 0, nrows-1
1.59 + ii = ind(cstrings(i,:).eq.cdelim(0))
1.60 +;
1.61 +; Make sure there were delimiters on this row. If not, we just quit.
1.62 +; This could probably be modified to do this more gracefully.
1.63 +;
1.64 + if(any(ismissing(ii))) then
1.65 + print("delim_indices: fatal: I didn't find any delimiters")
1.66 + print("('" + delimiter + "') on row " + i + ". Can't continue.")
1.67 + exit
1.68 + end if
1.69 + if(dimsizes(ii).ne.ndelims) then
1.70 + print("delim_indices: fatal: I expected to find " + ndelims)
1.71 + print("delimiters on row " + i + ". Instead, I found " + dimsizes(ii) + ".")
1.72 + print("Can't continue.")
1.73 + exit
1.74 + end if
1.75 +
1.76 + cindices(i,:) = ii
1.77 +
1.78 + delete(ii) ; For next time through loop
1.79 + end do
1.80 +
1.81 + return(cindices)
1.82 +end
1.83 +
1.84 +;----------------------------------------------------------------------
1.85 +; This function reads in a particular field from a string array,
1.86 +; given the field number to read (fields start at #1 and go to #nfield),
1.87 +; and the indices of the delimiters.
1.88 +;
1.89 +; It returns either an integer, float, double, character, or a string,
1.90 +; depending on the input flag "return_type".
1.91 +;----------------------------------------------------------------------
1.92 +function read_field(strings,ifield,indices,return_type)
1.93 +local nstring, cstrings, nf, tmp_str
1.94 +begin
1.95 + nrows = dimsizes(strings)
1.96 +;
1.97 +; Handle special case if we only have one string. Make sure it
1.98 +; is put into a 2D array.
1.99 +;
1.100 + if(nrows.eq.1) then
1.101 + cstrings = new((/1,strlen(strings)+1/),character)
1.102 + end if
1.103 +
1.104 + cstrings = stringtochar(strings)
1.105 + nf = dimsizes(indices(0,:))+1 ; indices is nrows x (nfields-1)
1.106 +
1.107 +;
1.108 +; Error checking. Make sure user has entered a valid field.
1.109 +;
1.110 + if(ifield.le.0.or.ifield.gt.nf) then
1.111 + print("read_field: fatal: you've selected a field that is")
1.112 + print("out-of-range of the number of fields that you have (" + nf + ").")
1.113 + exit
1.114 + end if
1.115 +
1.116 +;
1.117 +; Set up array to return. For string, int, float, or double arrays,
1.118 +; we don't have to do anything special. For character arrays,
1.119 +; however, we do.
1.120 +;
1.121 + if(return_type.ne."character") then
1.122 + return_array = new(nrows,return_type)
1.123 + else
1.124 +;
1.125 +; We don't know what the biggest character array is at this point, so
1.126 +; make it bigger than necessary, and then resize later as necessary.
1.127 +;
1.128 + tmp_return_array = new((/nrows,dimsizes(cstrings(0,:))/),"character")
1.129 +
1.130 + max_len = 0 ; Use to keep track of max lengths of strings.
1.131 + end if
1.132 +
1.133 + do i = 0,nrows-1
1.134 +;
1.135 +; Special case of first field in row.
1.136 +;
1.137 + if(ifield.eq.1) then
1.138 + ibeg = 0
1.139 + iend = indices(i,ifield-1)-1
1.140 + else
1.141 +;
1.142 +; Special case of first field in row.
1.143 +;
1.144 + if(ifield.eq.nf) then
1.145 + ibeg = indices(i,ifield-2)+1
1.146 + iend = dimsizes(cstrings(i,:))-1
1.147 +;
1.148 +; Any field between first and last field.
1.149 +;
1.150 + else
1.151 + ibeg = indices(i,ifield-2)+1
1.152 + iend = indices(i,ifield-1)-1
1.153 + end if
1.154 + end if
1.155 +;
1.156 +; Here's the code that pulls off the correct string, and converts it
1.157 +; to float if desired.
1.158 +;
1.159 + if(return_type.eq."integer") then
1.160 + return_array(i) = stringtointeger(chartostring(cstrings(i,ibeg:iend)))
1.161 + end if
1.162 + if(return_type.eq."float") then
1.163 + return_array(i) = stringtofloat(chartostring(cstrings(i,ibeg:iend)))
1.164 + end if
1.165 + if(return_type.eq."double") then
1.166 + return_array(i) = stringtodouble(chartostring(cstrings(i,ibeg:iend)))
1.167 + end if
1.168 + if(return_type.eq."string") then
1.169 + return_array(i) = chartostring(cstrings(i,ibeg:iend))
1.170 + end if
1.171 + if(return_type.eq."character") then
1.172 + if( (iend-ibeg+1) .gt. max_len) then
1.173 + max_len = iend-ibeg+1
1.174 + end if
1.175 + tmp_return_array(i,0:iend-ibeg) = cstrings(i,ibeg:iend)
1.176 + end if
1.177 + end do
1.178 +
1.179 + if(return_type.eq."character") then
1.180 + return_array = new((/nrows,max_len/),"character")
1.181 + return_array = tmp_return_array(:,0:max_len-1)
1.182 + end if
1.183 +
1.184 + return(return_array)
1.185 +end
1.186 +
1.187 +
1.188 +;----------------------------------------------------------------------
1.189 +; This function reads in string fields only to get the maximum string
1.190 +; length.
1.191 +;----------------------------------------------------------------------
1.192 +function get_maxlen(strings,ifield,indices)
1.193 +local nstring, cstrings, nf, tmp_str
1.194 +begin
1.195 + nrows = dimsizes(strings)
1.196 +;
1.197 +; Handle special case if we only have one string. Make sure it
1.198 +; is put into a 2D array.
1.199 +;
1.200 + if(nrows.eq.1) then
1.201 + cstrings = new((/1,strlen(strings)+1/),character)
1.202 + end if
1.203 +
1.204 + cstrings = stringtochar(strings)
1.205 + nf = dimsizes(indices(0,:))+1 ; indices is nrows x (nfields-1)
1.206 +
1.207 +;
1.208 +; Error checking. Make sure user has entered a valid field.
1.209 +;
1.210 + if(ifield.le.0.or.ifield.gt.nf) then
1.211 + print("read_field: fatal: you've selected a field that is")
1.212 + print("out-of-range of the number of fields that you have (" + nf + ").")
1.213 + exit
1.214 + end if
1.215 +;
1.216 +; We don't know what the biggest character array is at this point, so
1.217 +; make it bigger than necessary, and then resize later as necessary.
1.218 +;
1.219 + tmp_return_array = new((/nrows,dimsizes(cstrings(0,:))/),"character")
1.220 +
1.221 + max_len = 0 ; Use to keep track of max lengths of strings.
1.222 +
1.223 + do i = 0,nrows-1
1.224 +;
1.225 +; Special case of first field in row.
1.226 +;
1.227 + if(ifield.eq.1) then
1.228 + ibeg = 0
1.229 + iend = indices(i,ifield-1)-1
1.230 + else
1.231 +;
1.232 +; Special case of first field in row.
1.233 +;
1.234 + if(ifield.eq.nf) then
1.235 + ibeg = indices(i,ifield-2)+1
1.236 + iend = dimsizes(cstrings(i,:))-1
1.237 +;
1.238 +; Any field between first and last field.
1.239 +;
1.240 + else
1.241 + ibeg = indices(i,ifield-2)+1
1.242 + iend = indices(i,ifield-1)-1
1.243 + end if
1.244 + end if
1.245 + if( (iend-ibeg+1) .gt. max_len) then
1.246 + max_len = iend-ibeg+1
1.247 + end if
1.248 + end do
1.249 +
1.250 + return(max_len)
1.251 +end
1.252 +
1.253 +;----------------------------------------------------------------------
1.254 +; Main code.
1.255 +;----------------------------------------------------------------------
1.256 +begin
1.257 +
1.258 +; Set up defaults here. We are hard-coding the field types here.
1.259 +; You can set up this script to try to determine the field types
1.260 +; automatically, but this is a bit tedious. Maybe later.
1.261 +
1.262 + station = "USHa1"
1.263 + lat = 42.5378
1.264 + lon = -72.1715 + 360.
1.265 + year = 1991
1.266 + nfields = 30 ; # of fields
1.267 + delimiter = "," ; field delimiter
1.268 +
1.269 + filename = station+year+"_L4_m.txt" ; ASCII" file to read.
1.270 + cdf_file = station+year+"_L4_m.nc" ; netCDF file to write.
1.271 +
1.272 +; In this case, fields #2-#2 are integers,
1.273 +; and the rest of the fields are floats.
1.274 +
1.275 + var_types = new(nfields,string)
1.276 + var_strlens = new(nfields,integer) ; var to hold strlens, just in case.
1.277 +
1.278 + var_types = "float" ; Most are floats.
1.279 + var_types(0:1) = "integer"
1.280 +
1.281 + if(isfilepresent(cdf_file))
1.282 + print("Warning: '" + cdf_file + "' exists. Will remove it.")
1.283 + system("/bin/rm " + cdf_file)
1.284 + end if
1.285 +
1.286 +; Read in data as strings. This will create a string array that has the
1.287 +; same number of strings as there are rows in the file. We will then need
1.288 +; to parse each string later.
1.289 +
1.290 + read_data = asciiread(filename,-1,"string")
1.291 + header = read_data(0) ; Header. Use for variable names.
1.292 + data = read_data(1:) ; Get rid of first line which is a header.
1.293 + nrows = dimsizes(data) ; Number of rows.
1.294 +
1.295 +; Read in locations of delimiters in each string row.
1.296 +
1.297 + hindices = delim_indices(header,nfields,delimiter) ; header row
1.298 + dindices = delim_indices(data,nfields,delimiter) ; rest of file
1.299 +
1.300 +; print (hindices)
1.301 +; print (dindices)
1.302 +
1.303 +; Read in the field names which will become variable names on
1.304 +; the netCDF file.
1.305 +
1.306 + var_names = new(nfields,string)
1.307 +
1.308 + do i=0,nfields-1
1.309 + var_names(i) = read_field(header,i+1,hindices,"string")
1.310 + end do
1.311 +
1.312 +; Write out this netCDF file efficiently so it will be faster.
1.313 +; Try to predefine everything before you write to it.
1.314 +
1.315 + f = addfile(cdf_file,"c")
1.316 + setfileoption(f,"DefineMode",True) ; Enter predefine phase.
1.317 +
1.318 +; Write global attributes to file. It's okay to do this before
1.319 +; predefining the file's variables. We are still in "define" mode.
1.320 +
1.321 + fAtt = True
1.322 + fAtt@description = "Data read in from " + filename + " ASCII file."
1.323 + fAtt@creation_date = systemfunc ("date")
1.324 + fileattdef( f, fAtt )
1.325 +
1.326 +; Write dimension names to file. If there are no character variables,
1.327 +; then there's only one dimension name ("nvalues").
1.328 +
1.329 + nyear = -1
1.330 + nmonth = 12
1.331 +
1.332 + dim_names = (/ "year", "month" /)
1.333 + dim_sizes = (/ nyear , nmonth /)
1.334 + dimUnlim = (/ True , False /)
1.335 +
1.336 + filedimdef( f, dim_names, dim_sizes, dimUnlim )
1.337 + filedimdef( f, "lat", 1, False )
1.338 + filedimdef( f, "lon", 1, False )
1.339 +
1.340 +; Define each variable on the file.
1.341 +
1.342 + filevardef( f, "year", "integer", "year" )
1.343 + filevardef( f, "lat" , "float" , "lat" )
1.344 + filevardef( f, "lon" , "float" , "lon" )
1.345 +
1.346 +; Don't deal with variable Month (i=0).
1.347 +
1.348 + do i=0,nfields-1
1.349 + filevardef(f, var_names(i), var_types(i), dim_names)
1.350 + end do
1.351 +
1.352 +; Loop through each field, read the values for that field, print
1.353 +; information about the variable, and then write it to the netCDF
1.354 +; file.
1.355 +
1.356 + do i=0,nfields-1
1.357 + ifield = i+1 ; Fields start at #1, not #0.
1.358 +
1.359 +; Note: you can't write strings to a netCDF file, so these have
1.360 +; to be written out as character arrays.
1.361 +
1.362 + tmp_data = new((/1,nmonth/),var_types(i))
1.363 +
1.364 + tmp_data(0,:) = read_field(data,ifield,dindices,var_types(i))
1.365 +
1.366 +; tmp_data(0,:) = out_data(:)
1.367 +
1.368 +; Print some info about the variable.
1.369 +
1.370 + print("")
1.371 + print("Writing variable '" + var_names(i) + "' (field #" + ifield + ").")
1.372 + print("Type is " + var_types(i) + ".")
1.373 + print("min/max = " + min(tmp_data) + "/" + max(tmp_data))
1.374 +
1.375 + if(any(ismissing(tmp_data))) then
1.376 + print("This variable does contain missing values.")
1.377 + else
1.378 + print("This variable doesn't contain missing values.")
1.379 + end if
1.380 +
1.381 + f->$var_names(i)$ = tmp_data ; Write to netCDF file.
1.382 +
1.383 + delete(tmp_data) ; Delete for next round.
1.384 +; delete(out_data)
1.385 + end do
1.386 +
1.387 + f->year = year
1.388 + f->lat = lat
1.389 + f->lon = lon
1.390 +end