CESIUM内置shader变量和函数[转]

Cesium中内置了一些常量、变量和函数，在vs和fs中可直接使用。

内置uniform

内置uniform主要置于AutomaticUniforms类里面，该类私有未开放文档。

• czm_backgroundColor

An automatic GLSL uniform representing the current scene background color.
Example:
// GLSL declaration uniform vec4 czm_backgroundColor; // Example: If the given color's RGB matches the background color, invert it. vec4 adjustColorForContrast(vec4 color) { if (czm_backgroundColor.rgb == color.rgb) { color.rgb = vec3(1.0) - color.rgb; } return color; }

• czm_brdfLut

An automatic GLSL uniform containing the BRDF look up texture used for image-based lighting computations.
Example:
// GLSL declaration uniform sampler2D czm_brdfLut; // Example: For a given roughness and NdotV value, find the material's BRDF information in the red and green channels float roughness = 0.5; float NdotV = dot(normal, view); vec2 brdfLut = texture2D(czm_brdfLut, vec2(NdotV, 1.0 - roughness)).rg;

• czm_currentFrustum

An automatic GLSL uniform containing the near distance ( x) and the far distance ( y) of the frustum defined by the camera. This is the individual frustum used for multi-frustum rendering.
Example:
// GLSL declaration uniform vec2 czm_currentFrustum; // Example float frustumLength = czm_currentFrustum.y - czm_currentFrustum.x;

• czm_encodedCameraPositionMCHigh

An automatic GLSL uniform representing the high bits of the camera position in model coordinates. This is used for GPU RTE to eliminate jittering artifacts when rendering as described in Precisions, Precisions.
Example:
// GLSL declaration uniform vec3 czm_encodedCameraPositionMCHigh;

• czm_encodedCameraPositionMCLow

An automatic GLSL uniform representing the low bits of the camera position in model coordinates. This is used for GPU RTE to eliminate jittering artifacts when rendering as described in Precisions, Precisions.
Example:
// GLSL declaration uniform vec3 czm_encodedCameraPositionMCLow;

• czm_entireFrustum

An automatic GLSL uniform containing the near distance ( x) and the far distance ( y) of the frustum defined by the camera. This is the largest possible frustum, not an individual frustum used for multi-frustum rendering.
Example:
// GLSL declaration uniform vec2 czm_entireFrustum; // Example float frustumLength = czm_entireFrustum.y - czm_entireFrustum.x;

• czm_environmentMap

An automatic GLSL uniform containing the environment map used within the scene.
Example:
// GLSL declaration uniform samplerCube czm_environmentMap; // Example: Create a perfect reflection of the environment map on a model float reflected = reflect(view, normal); vec4 reflectedColor = textureCube(czm_environmentMap, reflected);

• czm_eyeHeight2D

An automatic GLSL uniform containing height ( x) and height squared ( y) of the eye (camera) in the 2D scene in meters.

• czm_fogDensity

An automatic GLSL uniform scalar used to mix a color with the fog color based on the distance to the camera.

• czm_frameNumber

An automatic GLSL uniform representing the frame number. This uniform is automatically incremented every frame.

• czm_frustumPlanes

The distances to the frustum planes. The top, bottom, left and right distances are the x, y, z, and w components, respectively.

• czm_geometricToleranceOverMeter

An automatic GLSL uniform scalar representing the geometric tolerance per meter

• czm_imagerySplitPosition

An automatic GLSL uniform representing the splitter position to use when rendering imagery layers with a splitter. This will be in pixel coordinates relative to the canvas.

• czm_infiniteProjection

An automatic GLSL uniform representing a 4x4 projection transformation matrix with the far plane at infinity, that transforms eye coordinates to clip coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output. An infinite far plane is used in algorithms like shadow volumes and GPU ray casting with proxy geometry to ensure that triangles are not clipped by the far plane.
Example:
// GLSL declaration uniform mat4 czm_infiniteProjection; // Example gl_Position = czm_infiniteProjection * eyePosition;

• czm_inverseModel

An automatic GLSL uniform representing a 4x4 model transformation matrix that transforms world coordinates to model coordinates.
Example:
// GLSL declaration uniform mat4 czm_inverseModel; // Example vec4 modelPosition = czm_inverseModel * worldPosition;

• czm_inverseModelView

An automatic GLSL uniform representing a 4x4 transformation matrix that transforms from eye coordinates to model coordinates.
Example:
// GLSL declaration uniform mat4 czm_inverseModelView; // Example vec4 modelPosition = czm_inverseModelView * eyePosition;

• czm_inverseModelView3D

An automatic GLSL uniform representing a 4x4 transformation matrix that transforms from eye coordinates to 3D model coordinates. In 3D mode, this is identical to czm_inverseModelView, but in 2D and Columbus View it represents the inverse model-view matrix as if the camera were at an equivalent location in 3D mode. This is useful for lighting 2D and Columbus View in the same way that 3D is lit.
Example:
// GLSL declaration uniform mat4 czm_inverseModelView3D; // Example vec4 modelPosition = czm_inverseModelView3D * eyePosition;

• czm_inverseModelViewProjection

An automatic GLSL uniform representing a 4x4 inverse model-view-projection transformation matrix that transforms clip coordinates to model coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output.
Example:
// GLSL declaration uniform mat4 czm_inverseModelViewProjection; // Example vec4 modelPosition = czm_inverseModelViewProjection * clipPosition;

• czm_inverseNormal

An automatic GLSL uniform representing a 3x3 normal transformation matrix that transforms normal vectors in eye coordinates to model coordinates. This is the opposite of the transform provided by czm_normal.
Example:
// GLSL declaration uniform mat3 czm_inverseNormal; // Example vec3 normalMC = czm_inverseNormal * normalEC;

• czm_inverseNormal3D

An automatic GLSL uniform representing a 3x3 normal transformation matrix that transforms normal vectors in eye coordinates to 3D model coordinates. This is the opposite of the transform provided by czm_normal. In 3D mode, this is identical to czm_inverseNormal, but in 2D and Columbus View it represents the inverse normal transformation matrix as if the camera were at an equivalent location in 3D mode. This is useful for lighting 2D and Columbus View in the same way that 3D is lit.
Example:
// GLSL declaration uniform mat3 czm_inverseNormal3D; // Example vec3 normalMC = czm_inverseNormal3D * normalEC;

• czm_inverseProjection

An automatic GLSL uniform representing a 4x4 inverse projection transformation matrix that transforms from clip coordinates to eye coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output.
Example:
// GLSL declaration uniform mat4 czm_inverseProjection; // Example vec4 eyePosition = czm_inverseProjection * clipPosition;

• czm_inverseView

An automatic GLSL uniform representing a 4x4 transformation matrix that transforms from eye coordinates to world coordinates.
Example:
// GLSL declaration uniform mat4 czm_inverseView; // Example vec4 worldPosition = czm_inverseView * eyePosition;

• czm_inverseView3D

An automatic GLSL uniform representing a 4x4 transformation matrix that transforms from 3D eye coordinates to world coordinates. In 3D mode, this is identical to czm_inverseView, but in 2D and Columbus View it represents the inverse view matrix as if the camera were at an equivalent location in 3D mode. This is useful for lighting 2D and Columbus View in the same way that 3D is lit.
Example:
// GLSL declaration uniform mat4 czm_inverseView3D; // Example vec4 worldPosition = czm_inverseView3D * eyePosition;

• czm_inverseViewProjection

An automatic GLSL uniform representing a 4x4 view-projection transformation matrix that transforms clip coordinates to world coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output.
Example:
// GLSL declaration uniform mat4 czm_inverseViewProjection; // Example vec4 worldPosition = czm_inverseViewProjection * clipPosition;

• czm_inverseViewRotation

An automatic GLSL uniform representing a 3x3 rotation matrix that transforms vectors from eye coordinates to world coordinates.
Example:
// GLSL declaration uniform mat3 czm_inverseViewRotation; // Example vec4 worldVector = czm_inverseViewRotation * eyeVector;

• czm_inverseViewRotation3D

An automatic GLSL uniform representing a 3x3 rotation matrix that transforms vectors from 3D eye coordinates to world coordinates. In 3D mode, this is identical to czm_inverseViewRotation, but in 2D and Columbus View it represents the inverse view matrix as if the camera were at an equivalent location in 3D mode. This is useful for lighting 2D and Columbus View in the same way that 3D is lit.
Example:
// GLSL declaration uniform mat3 czm_inverseViewRotation3D; // Example vec4 worldVector = czm_inverseViewRotation3D * eyeVector;

• czm_invertClassificationColor

An automatic GLSL uniform that will be the highlight color of unclassified 3D Tiles.

• czm_log2FarPlusOne

An automatic GLSL uniform containing log2 of the far distance + 1.0. This is used when reversing log depth computations.

• czm_log2NearDistance

An automatic GLSL uniform containing log2 of the near distance. This is used when writing log depth in the fragment shader.

• czm_minimumDisableDepthTestDistance

An automatic GLSL uniform representing the distance from the camera at which to disable the depth test of billboards, labels and points to, for example, prevent clipping against terrain. When set to zero, the depth test should always be applied. When less than zero, the depth test should never be applied.

• czm_model

An automatic GLSL uniform representing a 4x4 model transformation matrix that transforms model coordinates to world coordinates.
Example:
// GLSL declaration uniform mat4 czm_model; // Example vec4 worldPosition = czm_model * modelPosition;

• czm_modelView

An automatic GLSL uniform representing a 4x4 model-view transformation matrix that transforms model coordinates to eye coordinates.

Positions should be transformed to eye coordinates using czm_modelView and normals should be transformed using czm_normal.
Example:
// GLSL declaration uniform mat4 czm_modelView; // Example vec4 eyePosition = czm_modelView * modelPosition; // The above is equivalent to, but more efficient than: vec4 eyePosition = czm_view * czm_model * modelPosition;

• czm_modelView3D

An automatic GLSL uniform representing a 4x4 model-view transformation matrix that transforms 3D model coordinates to eye coordinates. In 3D mode, this is identical to czm_modelView, but in 2D and Columbus View it represents the model-view matrix as if the camera were at an equivalent location in 3D mode. This is useful for lighting 2D and Columbus View in the same way that 3D is lit.

Positions should be transformed to eye coordinates using czm_modelView3D and normals should be transformed using czm_normal3D.
Example:
// GLSL declaration uniform mat4 czm_modelView3D; // Example vec4 eyePosition = czm_modelView3D * modelPosition; // The above is equivalent to, but more efficient than: vec4 eyePosition = czm_view3D * czm_model * modelPosition;

• czm_modelViewInfiniteProjection

An automatic GLSL uniform representing a 4x4 model-view-projection transformation matrix that transforms model coordinates to clip coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output. The projection matrix places the far plane at infinity. This is useful in algorithms like shadow volumes and GPU ray casting with proxy geometry to ensure that triangles are not clipped by the far plane.
Example:
// GLSL declaration uniform mat4 czm_modelViewInfiniteProjection; // Example vec4 gl_Position = czm_modelViewInfiniteProjection * modelPosition; // The above is equivalent to, but more efficient than: gl_Position = czm_infiniteProjection * czm_view * czm_model * modelPosition;

• czm_modelViewProjection

An automatic GLSL uniform representing a 4x4 model-view-projection transformation matrix that transforms model coordinates to clip coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output.
Example:
// GLSL declaration uniform mat4 czm_modelViewProjection; // Example vec4 gl_Position = czm_modelViewProjection * modelPosition; // The above is equivalent to, but more efficient than: gl_Position = czm_projection * czm_view * czm_model * modelPosition;

• czm_modelViewProjectionRelativeToEye

An automatic GLSL uniform representing a 4x4 model-view-projection transformation matrix that transforms model coordinates, relative to the eye, to clip coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output. This is used in conjunction with czm_translateRelativeToEye.
Example:
// GLSL declaration uniform mat4 czm_modelViewProjectionRelativeToEye; // Example attribute vec3 positionHigh; attribute vec3 positionLow; void main() { vec4 p = czm_translateRelativeToEye(positionHigh, positionLow); gl_Position = czm_modelViewProjectionRelativeToEye * p; }

• czm_modelViewRelativeToEye

An automatic GLSL uniform representing a 4x4 model-view transformation matrix that transforms model coordinates, relative to the eye, to eye coordinates. This is used in conjunction with czm_translateRelativeToEye.
Example:
// GLSL declaration uniform mat4 czm_modelViewRelativeToEye; // Example attribute vec3 positionHigh; attribute vec3 positionLow; void main() { vec4 p = czm_translateRelativeToEye(positionHigh, positionLow); gl_Position = czm_projection * (czm_modelViewRelativeToEye * p); }

• czm_moonDirectionEC

An automatic GLSL uniform representing the normalized direction to the moon in eye coordinates. This is commonly used for directional lighting computations.
Example:
// GLSL declaration uniform vec3 czm_moonDirectionEC; // Example float diffuse = max(dot(czm_moonDirectionEC, normalEC), 0.0);

• czm_morphTime

An automatic GLSL uniform representing the current morph transition time between 2D/Columbus View and 3D, with 0.0 being 2D or Columbus View and 1.0 being 3D.
Example:
// GLSL declaration uniform float czm_morphTime; // Example vec4 p = czm_columbusViewMorph(position2D, position3D, czm_morphTime);

• czm_normal

An automatic GLSL uniform representing a 3x3 normal transformation matrix that transforms normal vectors in model coordinates to eye coordinates.

Positions should be transformed to eye coordinates using czm_modelView and normals should be transformed using czm_normal.
Example:
// GLSL declaration uniform mat3 czm_normal; // Example vec3 eyeNormal = czm_normal * normal;

• czm_normal3D

An automatic GLSL uniform representing a 3x3 normal transformation matrix that transforms normal vectors in 3D model coordinates to eye coordinates. In 3D mode, this is identical to czm_normal, but in 2D and Columbus View it represents the normal transformation matrix as if the camera were at an equivalent location in 3D mode. This is useful for lighting 2D and Columbus View in the same way that 3D is lit.

Positions should be transformed to eye coordinates using czm_modelView3D and normals should be transformed using czm_normal3D.
Example:
// GLSL declaration uniform mat3 czm_normal3D; // Example vec3 eyeNormal = czm_normal3D * normal;

• czm_orthographicIn3D

An automatic GLSL uniform that indicates if the current camera is orthographic in 3D.

• czm_pass

An automatic GLSL uniform representing the current rendering pass.
Example:
// GLSL declaration uniform float czm_pass; // Example if ((czm_pass == czm_passTranslucent) && isOpaque()) { gl_Position *= 0.0; // Cull opaque geometry in the translucent pass }

• czm_projection

An automatic GLSL uniform representing a 4x4 projection transformation matrix that transforms eye coordinates to clip coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output.
Example:
// GLSL declaration uniform mat4 czm_projection; // Example gl_Position = czm_projection * eyePosition;

• czm_resolutionScale

An automatic GLSL uniform representing the ratio of canvas coordinate space to canvas pixel space.
Example:
uniform float czm_resolutionScale;

• czm_sceneMode

An automatic GLSL uniform representing the current SceneMode, expressed as a float.
Example:
// GLSL declaration uniform float czm_sceneMode; // Example if (czm_sceneMode == czm_sceneMode2D) { eyeHeightSq = czm_eyeHeight2D.y; }

• czm_sunDirectionEC

An automatic GLSL uniform representing the normalized direction to the sun in eye coordinates. This is commonly used for directional lighting computations.
Example:
// GLSL declaration uniform vec3 czm_sunDirectionEC; // Example float diffuse = max(dot(czm_sunDirectionEC, normalEC), 0.0);

• czm_sunDirectionWC

An automatic GLSL uniform representing the normalized direction to the sun in world coordinates. This is commonly used for directional lighting computations.
Example:
// GLSL declaration uniform vec3 czm_sunDirectionWC;

• czm_sunPositionColumbusView

An automatic GLSL uniform representing the sun position in Columbus view world coordinates.
Example:
// GLSL declaration uniform vec3 czm_sunPositionColumbusView;

• czm_sunPositionWC

An automatic GLSL uniform representing the sun position in world coordinates.
Example:
// GLSL declaration uniform vec3 czm_sunPositionWC;

• czm_temeToPseudoFixed

An automatic GLSL uniform representing a 3x3 rotation matrix that transforms from True Equator Mean Equinox (TEME) axes to the pseudo-fixed axes at the current scene time.
Example:
// GLSL declaration uniform mat3 czm_temeToPseudoFixed; // Example vec3 pseudoFixed = czm_temeToPseudoFixed * teme;

• czm_view

An automatic GLSL uniform representing a 4x4 view transformation matrix that transforms world coordinates to eye coordinates.
Example:
// GLSL declaration uniform mat4 czm_view; // Example vec4 eyePosition = czm_view * worldPosition;

• czm_view3D

An automatic GLSL uniform representing a 4x4 view transformation matrix that transforms 3D world coordinates to eye coordinates. In 3D mode, this is identical to czm_view, but in 2D and Columbus View it represents the view matrix as if the camera were at an equivalent location in 3D mode. This is useful for lighting 2D and Columbus View in the same way that 3D is lit.
Example:
// GLSL declaration uniform mat4 czm_view3D; // Example vec4 eyePosition3D = czm_view3D * worldPosition3D;

• czm_viewerPositionWC

An automatic GLSL uniform representing the position of the viewer (camera) in world coordinates.

• czm_viewport

An automatic GLSL uniform containing the viewport's x, y, width, and height properties in an vec4's x, y, z, and w components, respectively.
Example:
// GLSL declaration uniform vec4 czm_viewport; // Scale the window coordinate components to [0, 1] by dividing // by the viewport's width and height. vec2 v = gl_FragCoord.xy / czm_viewport.zw;

• czm_viewportOrthographic

An automatic GLSL uniform representing a 4x4 orthographic projection matrix that transforms window coordinates to clip coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output.

This transform is useful when a vertex shader inputs or manipulates window coordinates as done by BillboardCollection.

Do not confuse czm_viewportTransformation with czm_viewportOrthographic. The former transforms from normalized device coordinates to window coordinates; the later transforms from window coordinates to clip coordinates, and is often used to assign to gl_Position.
Example:
// GLSL declaration uniform mat4 czm_viewportOrthographic; // Example gl_Position = czm_viewportOrthographic * vec4(windowPosition, 0.0, 1.0);

• czm_viewportTransformation

An automatic GLSL uniform representing a 4x4 transformation matrix that transforms normalized device coordinates to window coordinates. The context's full viewport is used, and the depth range is assumed to be near = 0 and far = 1.

This transform is useful when there is a need to manipulate window coordinates in a vertex shader as done by BillboardCollection. In many cases, this matrix will not be used directly; instead, czm_modelToWindowCoordinates will be used to transform directly from model to window coordinates.

Do not confuse czm_viewportTransformation with czm_viewportOrthographic. The former transforms from normalized device coordinates to window coordinates; the later transforms from window coordinates to clip coordinates, and is often used to assign to gl_Position.
Example:
// GLSL declaration uniform mat4 czm_viewportTransformation; // Use czm_viewportTransformation as part of the // transform from model to window coordinates. vec4 q = czm_modelViewProjection * positionMC; // model to clip coordinates q.xyz /= q.w; // clip to normalized device coordinates (ndc) q.xyz = (czm_viewportTransformation * vec4(q.xyz, 1.0)).xyz; // ndc to window coordinates

• czm_viewProjection

An automatic GLSL uniform representing a 4x4 view-projection transformation matrix that transforms world coordinates to clip coordinates. Clip coordinates is the coordinate system for a vertex shader's gl_Position output.
Example:
// GLSL declaration uniform mat4 czm_viewProjection; // Example vec4 gl_Position = czm_viewProjection * czm_model * modelPosition; // The above is equivalent to, but more efficient than: gl_Position = czm_projection * czm_view * czm_model * modelPosition;

• czm_viewRotation

An automatic GLSL uniform representing a 3x3 view rotation matrix that transforms vectors in world coordinates to eye coordinates.
Example:
// GLSL declaration uniform mat3 czm_viewRotation; // Example vec3 eyeVector = czm_viewRotation * worldVector;

• czm_viewRotation3D

An automatic GLSL uniform representing a 3x3 view rotation matrix that transforms vectors in 3D world coordinates to eye coordinates. In 3D mode, this is identical to czm_viewRotation, but in 2D and Columbus View it represents the view matrix as if the camera were at an equivalent location in 3D mode. This is useful for lighting 2D and Columbus View in the same way that 3D is lit.
Example:
// GLSL declaration uniform mat3 czm_viewRotation3D; // Example vec3 eyeVector = czm_viewRotation3D * worldVector;

内置常量

这里常量和函数的定义，在cesium官网的一个历史文档里有描述，后来版本的文档里没有了。

https://cesiumjs.org/releases/b28/Documentation/index.html

点击glsl，可以看到。

• czm_degreesPerRadian

• czm_depthRange

• czm_epsilon1

• czm_epsilon2

• czm_epsilon3

• czm_epsilon4

• czm_epsilon5

• czm_epsilon6

• czm_epsilon7

• czm_infinity

• czm_oneOverPi

• czm_oneOverTwoPi

• czm_passCesium3DTile

• czm_passCesium3DTileClassification

• czm_passCesium3DTileClassificationIgnoreShow

• czm_passClassification

• czm_passCompute

• czm_passEnvironment

• czm_passGlobe

• czm_passOpaque

• czm_passOverlay

• czm_passTerrainClassification

• czm_passTranslucent

• czm_pi

• czm_piOverFour

• czm_piOverSix

• czm_piOverThree

• czm_piOverTwo

• czm_radiansPerDegree

• czm_sceneMode

• czm_sceneMode2D

• czm_sceneMode3D

• czm_sceneModeColumbusView

• czm_sceneModeMorphing

• czm_solarRadius

• czm_threePiOver2

• czm_twoPi

• czm_webMercatorMaxLatitude

内置结构体定义

• czm_depthRangeStruct

• czm_ellipsoid

• czm_material

• czm_materialInput

• czm_ray

• czm_raySegment

• czm_shadowParameters

内置函数

czm_alphaWeight

• czm_antialias

• czm_approximateSphericalCoordinates

• czm_branchFreeTernary

• czm_cascadeColor

• czm_cascadeDistance

• czm_cascadeMatrix

• czm_cascadeWeights

• czm_columbusViewMorph

• czm_computePosition

• czm_cosineAndSine

• czm_decompressTextureCoordinates

• czm_depthClampFarPlane

• czm_eastNorthUpToEyeCoordinates

• czm_ellipsoidContainsPoint

• czm_ellipsoidNew

• czm_ellipsoidWgs84TextureCoordinates

• czm_equalsEpsilon

• czm_eyeOffset

• czm_eyeToWindowCoordinates

• czm_fastApproximateAtan

• czm_fog

• czm_gammaCorrect

• czm_geodeticSurfaceNormal

• czm_getDefaultMaterial

• czm_getLambertDiffuse

• czm_getSpecular

• czm_getWaterNoise

• czm_getWgs84EllipsoidEC

• czm_HSBToRGB

• czm_HSLToRGB

• czm_hue

• czm_inverseGamma

• czm_isEmpty

• czm_isFull

• czm_latitudeToWebMercatorFraction

• czm_lineDistance

• czm_luminance

• czm_metersPerPixel

• czm_modelToWindowCoordinates

• czm_multiplyWithColorBalance

• czm_nearFarScalar

• czm_octDecode

• czm_packDepth

• czm_phong

• czm_planeDistance

• czm_pointAlongRay

• czm_rayEllipsoidIntersectionInterval

• czm_readDepth

• czm_reverseLogDepth

• czm_RGBToHSB

• czm_RGBToHSL

• czm_RGBToXYZ

• czm_sampleOctahedralProjection

• czm_saturation

• czm_shadowDepthCompare

• czm_shadowVisibility

• czm_signNotZero

• czm_sphericalHarmonics

• czm_tangentToEyeSpaceMatrix

• czm_transformPlane

• czm_translateRelativeToEye

• czm_translucentPhong

• czm_transpose

• czm_unpackDepth

• czm_unpackFloat

• czm_vertexLogDepth

• czm_windowToEyeCoordinates

• czm_writeDepthClampedToFarPlane

• czm_writeLogDepth

• czm_XYZToRGB

• czm_emptyRaySegment

• czm_fullRaySegment

原文：https://www.cnblogs.com/wanghui2011/articles/10870294.html