In this textbook, I found the following remark:
An $(n,k)$ linear code $\mathcal{C}$ is a unique subspace consisting of a set of $2^k$ codewords.
The statement surprised me because in vector spaces over infinite fields like $\mathbb{R}^n$, there are infinitely many subspaces with dimension $k<n$.
Below is my attempt to prove the statement.
Let $\mathcal{C}_1$ and $\mathcal{C_2}$ be $(n,k)$ linear codes. Consider their generator matrices in systematic form $\mathbf{G}_1=[\mathbf{I}_k|\mathbf{P}_1]$ and $\mathbf{G}_2=[\mathbf{I}_k|\mathbf{P}_2]$. By symmetry, it suffices to show that every codeword in $\mathcal{C}_1$ is a codeword in $\mathcal{C}_2$.
Let $\mathbf{u}$ be a binary $1\times k$ vector. Then, the corresponding codeword in $\mathcal{C}_1$ is $\mathbf{x}=[\mathbf{u}|\mathbf{u}\mathbf{P}_1]$. Then, a necessary condition so that $\mathbf{x}\in\mathcal{C}_2$ is to set $\mathbf{u}=\mathbf{v}$ so that
$$ \mathbf{x}=[\mathbf{u}|\mathbf{u}\mathbf{P}_1]=[\mathbf{v}|\mathbf{v}\mathbf{P}_2], $$
Then, to complete the proof, I have to show that $\mathbf{u}(\mathbf{P}_1-\mathbf{P}_2)=\mathbf{0}$.
My question is:
What argument can I use to prove that $\mathbf{P}_1 =\mathbf{P}_2$ using what I have right now?
A $(n,k)$ linear code $\mathcal{C}$ is a subspace living in the vector space $(\mathbb{F}^n,\mathbb{F}_2,+_2,.)$ which has $2^k$ distinct code words, is the right way to read it.
Let $G$ be the generator matrix for $\mathcal{C}$. Since $\mathcal{C}$ is a $k$ dimensional subspace in $\mathbb{F}^n_2$, $G$ is a $k \times n$. Then for any vector $x=(x_1,x_2\dots,x_k), \hspace{0.2cm}$ $ x_i\in \mathbb{F}_2$, $G^Tx$ is a codeword. Suppose for some $x_1,x_2$, $G^Tx_1=G^Tx_2$ then, $G^T(x_1-x_2)=0$. This implies $x_1-x_2 \in \text{ Nullspace }(G^T)$. But $G^{T}$ is a matrix with linearly independent columns by definition and hence only the zero linear combination fetch the $0$ codeword and thus, $x_1-x_2=0$, guaranteeing uniqness of codewords. Finally as there are $2^k$ possible $x$ vectors, the statement holds true.