Python NetCDF IOError: netcdf: 无效的维度ID或名称

Question

我正在用Python写一个处理NetCDF文件的脚本，但在创建变量时遇到了一些问题，这里是代码的一部分：

stepnumber_var = ofl.createVariable("step_number", "i",("step_number",))
stepnumber_var.standard_name = "step_number"

atomNumber_var = ofl.createVariable("atom_number", "i", ("atom_number",))
atomNumber_var.standard_name = "atom__number"

但是出现了这个错误：

Traceback (most recent call last):
  File "sub_avg.py", line 141, in <module>
    atomNumber_var = ofl.createVariable("atom_number", "i", ("atom_number",))
IOError: netcdf: NetCDF: Invalid dimension ID or name

我的问题是，为什么第一个变量可以正常创建，而第二个变量却不行呢？

谢谢

这是完整的代码：

from array import array 
import os
import sys
import math
import string as st
import numpy as N
from Scientific.IO.NetCDF import NetCDFFile as S

if len(sys.argv) < 2:
    sys.exit( "No input file found. \nPlease privide NetCDF trajectory input file" )
#######################
## Open NetCDF file ### 
#######################
infl = S(sys.argv[1], 'r')  

file = sys.argv[1]
title,ext = file.split(".")

                                #for v in infl.variables:   # Lists the variables in file
                                #   print(v)        

#################################################################################
# Variable "configurations" has the structure [step_number, atom_number, x y z] #
#################################################################################

varShape = infl.variables['configuration'].shape        # This gets the shape of the variable, i.e. the dimension in terms of elements

nSteps = varShape[0]                                
nAtoms = varShape[1]


coordX_atom = N.zeros((nSteps,nAtoms))
coordY_atom = N.zeros((nSteps,nAtoms))
coordZ_atom = N.zeros((nSteps,nAtoms))

sumX = [0] * nAtoms
sumY = [0] * nAtoms
sumZ = [0] * nAtoms

######################################################
# 1) Calculate the average structure fron trajectory #
######################################################

for i in range(0, 3):
    for j in range(0, 3):
        coordX_atom[i][j] = infl.variables["configuration"][i,j,0]
        coordY_atom[i][j] = infl.variables["configuration"][i,j,1]
        coordZ_atom[i][j] = infl.variables["configuration"][i,j,2]

        sumX[j] = sumX[j] + coordX_atom[i][j]
        sumY[j] = sumY[j] + coordY_atom[i][j]
        sumZ[j] = sumZ[j] + coordZ_atom[i][j]

avgX = [0] * nAtoms
avgY = [0] * nAtoms
avgZ = [0] * nAtoms

for j in range(0, 3):
    avgX[j] = sumX[j]/nSteps 
        avgY[j] = sumY[j]/nSteps
        avgZ[j] = sumZ[j]/nSteps

##############################################################
# 2) Subtract average structure to each atom and for each frame #
##############################################################

for i in range(0, 3):
    for j in range(0, 3):
                coordX_atom[i][j] = infl.variables["configuration"][i,j,0] - avgX[j]
                coordY_atom[i][j] = infl.variables["configuration"][i,j,1] - avgY[j]
                coordZ_atom[i][j] = infl.variables["configuration"][i,j,2] - avgZ[j]

#######################################
# 3) Write new NetCDF trajectory file #                      
#######################################

ofl = S(title + "_subAVG.nc", "a")
############################################################
# Get information of variables contained in the NetCDF input file
#############################################################

i = 0
for v in infl.variables:       
    varNames = [v for v in infl.variables]
    i += 1
#############################################
# Respectively get, elements names in variable, dimension of elements and lenght of the array variableNames
##############################################
for v in infl.variables["box_size"].dimensions:
    boxSizeNames = [v for v in infl.variables["box_size"].dimensions]
for v in infl.variables["box_size"].shape:
    boxSizeShape = [v for v in infl.variables["box_size"].shape]
boxSizeLenght = boxSizeNames.__len__()

print boxSizeLenght

for v in infl.variables["step"].dimensions:
    stepNames = [v for v in infl.variables["step"].dimensions]
for v in infl.variables["step"].shape:
    stepShape = [v for v in infl.variables["box_size"].shape]
stepLenght = stepNames.__len__()
print stepLenght

for v in infl.variables["configuration"].dimensions:
    configurationNames = [v for v in infl.variables["configuration"].dimensions]
for v in infl.variables["configuration"].shape:
    configurationShape = [v for v in infl.variables["configuration"].shape]
configurationLenght = configurationNames.__len__()
print configurationLenght

for v in infl.variables["description"].dimensions:
    descriptionNames = [v for v in infl.variables["description"].dimensions]
for v in infl.variables["description"].shape:
    descriptionShape = [v for v in infl.variables["description"].shape]
descriptionLenght = descriptionNames.__len__()
print descriptionLenght

for v in infl.variables["time"].dimensions:
    timeNames = [v for v in infl.variables["time"].dimensions]
for v in infl.variables["time"].shape:
    timeShape = [v for v in infl.variables["time"].shape]
timeLenght = timeNames.__len__()
print timeLenght

#Get Box size

xBox =  infl.variables["box_size"][0,0]
yBox =  infl.variables["box_size"][0,1]
zBox =  infl.variables["box_size"][0,2]

# Get description lenght
description_lenghtLenght = infl.variables["description"][:]

############################################################
# Create Dimensions
############################################################

stepnumber_var = ofl.createVariable("step_number", "i",("step_number",))
stepnumber_var.standard_name = "step_number"

atomNumber_var = ofl.createVariable("atom_number", "i", ("atom_number",))
atomNumber_var.standard_name = "atom__number"


#
#xyz_var = ofl.createVariable("xyz", "f",("xyz",))
#xyz_var.units = "nanometers"
#xyz_var.standard_name = "xyz"
#
#configuration_var = ofl.createVariable("configuration", "f", ("step_number", "atom_number", "xyz"))
#configuration_var.units = "nanometers"
#configuration_var.standard_name = "configuration"
#
#print configuration_var.shape
#step_var = ofl.createVariable("box_size_lenght", 3)
#configuration_var = ofl.createVariable("atom_number", nAtoms)
#description_var = ofl.createVariable("xyz", 3)
#time_var = ofl.createVariable(description_lenght, description_lenghtLenght)
#
#a = infl.variables["step_number"].dimensions.keys()
#print a

谢谢！

Answer 1

atomNumber_var.standard_name = "atom__number"

这里的atom__number用了两个"__"，而不是一个"_"。我不确定这是不是你的问题，但可以考虑看看。

我还建议你把netcdf文件的步骤写得更清楚。我喜欢把它分成三步。我用的是一个关于海洋表面温度的科学数据的例子。你有一个创建维度的部分，但实际上并没有做到。这部分更准确来说应该是创建变量的部分。

1. 创建维度

2. 创建变量

3. 填充变量

   from netCDF4 import Dataset
   ncfile = Dataset('temp.nc','w')
   lonsdim = latdata.shape    #Set dimension lengths
   latsdim = londata.shape   
   ###############
   #Create Dimensions
   ###############
   latdim   = ncfile.createDimension('latitude', latsdim)
   londim   = ncfile.createDimension('longitude', lonsdim)
   ###############
   #Create Variables
   #################   The variables contain the dimensions previously set
   latitude  = ncfile.createVariable('latitude','f8',('latitude'))
   longitude = ncfile.createVariable('longitude','f8',('longitude'))
   oceantemp  = ncfile.createVariable('SST','f4' ('latitude','longitude'),fill_value=-99999.0)
   ###############
   Fill Variables
   ################
   latitude[:]    = latdata      #lat data to fill in
   longitude[:]   = londata      #lon data to fill in
   oceantemp[:,:]  = sst[:,:]    #some variable previous calculated
   

希望这些对你有帮助。

Answer 2

这可能是一个库在试图“帮忙”的情况（具体情况我在帖子最后会详细说明，但我不能确认）。要解决这个问题，你需要明确为 atom_number 和 step_number 创建维度，可以在创建变量之前使用以下代码（假设我对 nSteps 和 nAtoms 的理解是正确的）：

ofl.createDimension("step_number", nSteps)
ofl.createDimension("atom_number", nAtoms)

如果你对 netCDF 还不太熟悉，我建议你看看 netcdf4-python 这个包，

http://unidata.github.io/netcdf4-python/

或者在 scipy 中找到的 netCDF 包：

http://docs.scipy.org/doc/scipy/reference/io.html

可能发生的情况是：当你创建变量 step_number 时，库试图帮忙，创建了一个无限长度的 step_number 维度。然而，在 netcdf-3 文件中，你只能有一个无限维度，所以这个“帮忙”的做法并不奏效。